Aggregation improvements (hoesh)

Cool, I like this, while got some buf fixes, optimization, feature improvement:

1) Simple concatenation failed in my case, where a files EOF was just right after
a closing brace '}', and the next file was started with a statement. This could
cause the TinyMCE problem too, but not sure. If the explained situation happens,
the left of the aggregated JS file was discarded with syntax error.

2) I've put a comment into aggregated file with information regarding what was
aggregated. Makes inspection of js to be excluded easier.

3) Created the mentioned pattern matching compiler. Right now if you jut type ".js"
in the exclusion list, then all .js files will be excluded. Well, this is not the way you
employ this feature generally, but useful, if you dont want to type the whole path.

4) Made aggregated file name generation file modification sensitive. included files',
and the module's .module file's modification time are counted into md5 now.
This makes caching reliable when updating any module including the aggregator.

As I dont have dif & patch on my Mac right now I just attach the modified file.
(I have used Chirale's version) Please someone create a patch against HEAD in CVS,
rewiew the code (small optimization to the mail staff happened), and test it.

Thanks,

Endre