save const, but apparently they're not. *shrug*

&file_count is being passed as a pointer to a size_t.

This need to be tested.

Mingw32 is _WIN32 and GCC, and we need some way to get the free()
prototype. :D

The last element of these arrays was getting clobbered by the next
field because the array sizes are stored in the first element.

Fixes warning on NetBSD and Raspian.

It looks like we may need a set of PRI?TIME macros for printf()ing
a time_t.

Also, add a BUILD_ARGS variable so custom pipelines can be ran
for the templated job tree.

Just because I use them for SyncTERM releases doesn't mean we want
them for CI.

This double definition just seems confusing and error-prone.
Sure, the interpolation form is more verbose and ugly, but at least
it's not duplicative.

This utility is for manual testing, not automated testing, so there's
no point in anyone building it as a side-effect.

Removing it should also flatten the dependency tree and allow more
parallel jobs (there's up to 18 seconds to gain!)