Make RW lock work recursively

[Duttenheim]

Currently, the RW locks might not work recursively for some platforms. For example, The SRWLOCK in Win32 doesn't allow for reentrant acquiring of the lock even if on the same thread, and will cause a deadlock.

This should be implemented for all platforms.

One suggested method is to use thread local counters and only perform the lock and unlock when the counter is 0, but always increment the counter.