[patch v1] malloc: add malloc_zero_aborts tunable

[DJ Delorie]

Florian Weimer <fweimer@redhat.com> writes:
> We can't do that for static builds, yes.  I don't think that would be
> particularly worthwhile anyway.

I would be happy with static builds being stuck with default malloc ;-)

I really just wanted to get *anything* working.  I mean, the framework
works, and if I had something to use to decide which malloc to choose I
could have done that.  What I couldn't figure out how to do was to
reliably get a choice from the user that early in program loading.  Had
I more time, my next option was to dig out the original environment
pointer and store it somewhere internally where I could manually parse
it.

Most of the existing ifuncs use cpuid, although you can mask them with
tunables, but I couldn't figure out why those tunables worked and mine
didn't (other than I couldn't call tunables_get() that early, and they
call it later?)