the regex backend already returned 2 for an internal failure, contrary to
documentation, so this mainly brings the latter in line with existing behavior.

at the same time, fix a bug: h_compile would previously return "false" (0) on
exceptions (longjmp).

make use of our new freedom to signal different error conditions from the
context-free backends, notably "incompatible combinator" vs. "unresolvable
conflict".

We now have a `stopwatch` to measure time between two moments.

It's causing unreferenced-variable warnings, and isn't referenced
anywhere aside from benchmark.c. If client code is likely to reference
it, perhaps move it into another header, so people who include hammer.h
don't have to refer to it to have warning-free code.

CLOCK_THREAD_CPUTIME_ID is a linux-specific timer. On NetBSD, the best
way to get timer information appears to be getrusage, which happens to
be fairly cross-platform.

effects throughout, including use of inttypes.h/PRI[ud]64 because printf() is
still stuck in a long world, conversion of %lu to %zu for printing values of
type size_t, and changing/renaming the g_check_cmp* family of functions.

This reverts commit d81d820d.

Benchmark is actually broken right now, for mysterious double-free related reasons. I will dig into this tomorrow.