bug#36597: 27.0.50; rehash hash tables eagerly in pdumper

[Paul Eggert]

Pip Cet wrote:
> I don't see anything in there that
> would drastically affect performance or memory footprint.

The drastic change to the memory footprint occurs because the old code 
incorrectly computed the length of the hash table's vectors, and over-allocated 
them in some cases. The over-allocation factor could get worse with each hash 
table resize, following a Fibonacci-like sequence. I think this was a bug I 
introduced in 2011-07-21T17:41:20!eggert@cs.ucla.edu.