[PATCH] Fix tcache count maximum

[Wilco Dijkstra]

Hi DJ,

> Putting in an upper limit gives the user an answer, and since it's a
> small number, it need not be precise.

Sure, bit documenting this value means we might have to update it
when changing the internal implementation of tcache.

>> but the key point is that we can improve the tcache significantly.
>
> Please do!  I knew going into this that the tcache could have lots of
> enhancements, but I wanted to keep it simple for review purposes.
> 
> The only tricky part is agreeing on which benchmarks are valid ;-)

All in principle! jemalloc is gaining popularity so any case where it beats
GLIBC is a reason to improve it to stay competitive.

Wilco