[PATCH 09 of 10] subrepo: ensure "lock.release()" execution at the end of "_cachestorehash()"

[timeless at gmail.com]

> in ordinary use-cases, lock should be released via destructor of
> the "lock" class at the end of each "hg" command invocations and
> not so serious (shouldn't it ?)‎

I think CommandServer is long lived: http://mercurial.selenic.com/wiki/CommandServer