[PATCH] Making Mem::Init() re-entrant.

Moving the cache_peer pool to the peer init area has brought to light
the problem of this pool init function not being safe for repeated
calls. (bug 3245)

We had some discussion when the opposite of this came up about strings.

Since we have not really settled on whether to make MemPools an
on-first-demand system I think its time to make some kind of TODO plan now.

Do we want to eventually make memory initialization happen on first use?
   In which case I commit this patch now as one small step towards that.

Or do we want to set an absolute init order and fail loudly if
components are initialized in the wrong order?
  In which case I revert the peerSelectInit() changes.

Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE9 or 3.1.12
   Beta testers wanted for 3.2.0.9 and 3.1.12.3