* Fix bug where many elements in the code completion popup did not show up as private
(even though they are showing up as private in the navigation outline)

* Fix a bug where trying to ctrl-click for Go To Declaration into an item that comes from the
preindexed documentation (Yahoo, JQuery, Dojo) would give a misleading error message
(it would say "Foo is a built-in method and has no JavaScript source"). This fix displays
the more accurate "Cannot find the source for Foo". It also fixes an NPE that might occur
if the documentation for a function is not found.

* Fix a possible NPE which can happen when you're doing code completion on some index entries. The problem
was that you could end up with searching for both name=null,in=null which is invalid.

* Fix bug where code completion was showing keywords in many places they are not valid. For example,
doing completion like this:
x = new |
or
y.|
should not show keywords like while, for, do, etc. - they are clearly not allowed in these places.

Also, fix new completion (the first case above) when the caret is at the end of the line (right in
front of a \n/EOL).

* Fix bug where "new" completion (which lists possible constructors) would not include many classes
that have default constructors. These aren't stored as contructor methods in the index; they are
stored as properties, so update the index search to handle this.

* Fix a couple of bugs in indexing.
First, @attribute is synonymous with @property, and occurs in some of the YUI documentation.
This is covered by the updated unit test JsIndexerTest.testScriptDoc2

Also, fix a bug in extracting data from the sdoc.zip file; it was computing incorrect filename URLs
when recursing through the file structure. It should be using FileObject.getNameExt(), not FileObject.getName(),
when creating a URL; without this, the stored fileurls were missing their .sdoc extension and documents
couldn't be found.

* Fix bug in error parser recovery.
I based this for JavaScript from the similar error recovery in Ruby -- but I forgot to update the
comment delimiter from "#" to "//"...

* 131667: Overly aggressive "code has no side effects" warning
Fix the above bug: suppress warnings on Token.EMPTY nodes in HTML files. These are added into
the HTML file as part of the HTML to JavaScript embedding model creation (I add some extra semicolons
to prevent accidentally continuing statements from one <script> block to the next.).

This fix also corrects a bug in offset computation where the trailing comma quickfix would
show up on the wrong line in HTML files - because I had forgotten to do an AST to lexer based position
translation which always has to be done or things don't work in embedded files.

* Fix bug in the unsupported calls quickfix. It wasn't able to compute the call type for AST expressions
where you had nested function calls. This was because I was basing the hint on the Token.NAME node instead
of the correct Token.STRING node which is always present for calls; the Token.NAME node is only there
for fully-qualified function call names. I also changed the quickfix popup dialog title from listing
the bit long "Function {0} is not supported by all of your targeted browsers" which caused wide dialogs,
to just the fully qualified name of the function.

* The Trailing Comma warning should have severity error, not warning - since it's an error you
really need to deal with.

* Modify the Rhino AST for switch-statements. Rhino would rewrite the AST quite a bit where it no
longer reflects the source structure. This is bad for the IDE in a couple of ways. Martin
needs this fixed for the AST-based formatter.

* 131665: NPE in GSF
This is just a documentation fix to gsf's javadoc to clarify the usage of an API which was
called with invalid params from PHP.

* 131671: NPE in GSF
This is a fix for a scenario which has happened to some users where GSF tries to index a file
which is no longer there -- and the file is named something like ".#foo". These aren't files
we shouldn't try to index anyway - they are shortlived files typically created a backup files
by various editors and/or version control systems.


(Also update lots of unit tests for the above fixes)

Fix bug in GSF for computing quickfixes in embedded files

Fix CSS selector completion and HTML id/class completion for Prototype and JQuery; jsdoc tag completion in comments. Tests.

JavaScript browser compatibility, and GSF task cancellation for some source tasks

128985: Appearance of the HTML navigator

#128038 - html navigator should be divided into nodes according to embedded languages

JRuby 1.1RC3 update

* During startup indexing, look at the size of files and don't attempt
to open files whose sizes are > 1Mb. This same check is performed in openide,
so before this fix opening some projects would cause dialog popups asking
whether you really wanted to open this file (with only an OK button,
no cancel!)
* One of the files which caused this message was the YUI precomputed library
file. I redid the code generator for it to split it up into lots of smaller
files, one for each corresponding YUI js file, and adjusted the linking
code to deal with this scenario.
* Fix formatting bug such taht the JS Doc formatter more accurately detects
whether it should blockquote the comment or not.
* Support @name and @memberOf tags in the sdoc, which is used by the
Woodstock library

A bunch of JavaScript and GSF changes:
* Update the pre-index files. They were reverted to an earlier version by
the merging fiasco earlier this week. This fixes #130662.
* Support for sdoc, documentation specified separately from the javascript source
and automatically linked to it. This lets us provide documentation for libraries
like jQuery where the included library doesn't have any documentation that we can
extract. I'll probably switch the core stubs over to using this as well.
* GSF: Update the GsfLanguage and Indexer APIs to allow language implementations to
filter out classpaths that don't apply; for a mixed project like Rails for example
this lets JavaScript bypass all the Ruby Gem directories in index queries.
Also make display name and optional preferred extension be part of the language
configuration object rather than be registered in the layer. (The displayname
is mandatory and was missing from a couple of gsf clients.)
* Fix bug in language embedding which resulted in bogus JavaScript errors on
lots of HTML files in the YUI samples
* Switch to a declarative mime resolver for JavaScript instead of the hardcoded
one since it doesn't need any custom logic.
* Fix bug triggered by some JavaScript files where functions with "two"
names caused assertion (functions that are both labelled in an object
literal as well as specifying their own name; this is unusual)
* Remove filenames from the completion right-hand-side for packages and classes
since the filenames there were misleading
* No longer silently eat exceptions encountered during startup indexing; these are logged
* Automate preindexing scripts so no manual steps are needed
* Hopefully fix 128191: NullPointerException at org.netbeans.modules.gsf.GsfTaskProvider$1.taskFinished

Revert of #d54bd12bad21