A parsing proposal

[Jonas Nordin]

> For large projects (KDE) this will be a significant increase in
> performance.
> 
> I will add this to my Intellisense document.  I have been taking all the
> good ideas and writing them down so that we can have a robust
> intellisense implementation

Everything concerning the parsing of classes and their methods has already been
taken care of in the ClassParser class, this included persistent storage of
preparsed classes. The problem right now is to store the classes in a database. 

My first intention was to use gdbm, but as some users has pointed out gdbm
doesn't handle the situation with one-writer multiple readers. This means that
you only can start one instance of kdevelop. This is something we have to fix.

One more concern is that the lookup-time must be low, the intention is to
minimize memory usage. This means that I don't want to keep any object
references I memory when not necessary.

/J