Use precise=false for performance when parsing unchanging code

[redmunds]

This is for #4608 (please ignore branch name). See bug for details.

@njx @RaymondLim have been following this one.

Are there any other places where we can isolate calls to getTokenAt() and set precise=false?

[RaymondLim]

@redmunds Nice job in finding the culprit and the solution. I think we may also improve the performance in js files as we have a couple of getTokenAt() calls with precise=true in extensions/default/JavaScriptCodeHints/Session.js.

[redmunds]

@iwehrman This should also fix the performance problem you were seeing with parsing <style> blocks.

[redmunds]

@RaymondLim Good catch! I also made fix in extensions/default/JavaScriptCodeHints/Session.js .

[RaymondLim]

@redmunds Can you rebase your changes to the latest master or create a new branch with the latest master and reapply your changes? The code changes look good, but when I run your branch, I didn't see text as I type. It took a while to show up a block of text and then to show JS hints. So it's not clear your changes fix the issue or not.

[redmunds]

@RaymondLim There are only 2 commits that change different files, so there's nothing to rebase. I merged the latest code from master, but it was only 13 commits.

The main change should make the initial loading of JS faster, so first hint will display faster. After that it should have no effect. To test this, open 2 files (1 HTML with <script> block, and 1 JS file) and switch back and forth (hints get reloaded whenever you switch to different file). You may also see this improvement with Quick Open of both JS and CSS.

The second change I made for JS Code Hints is general parsing around cursor location. If you can't see a difference, then maybe it should be backed out to reduce risk.

[RaymondLim]

Looks good. Merging.