Problem: interrupting the solver is impossible

I'm hacking away at #51245. In particular see the followup tasks I'm working on at #51245 (comment), which note how the solver is unresponsive to ^C when it triggers the invalid variant condition described in that issue.

It turns that clingo isn't doing anything wrong, but we really need to avoid blocking on the SolveHandle.wait(-1) method. Without spack/spack-packages#1600, the following hangs:

; spack spec icu4c
# hangs forever
^C^C^C^C^C
^C^C^C
# unresponsive to ^C
# requires `kill -s TERM/KILL` from a separate terminal
; spack -d spec icu4c
# produces some debug output before initiating the solve,
# but no output after initiating the async solve

Aside: this is a really frustrating failure mode

I triggered this failure mode by accident trying to add new build variants for python 3.15 --disable-gil, and I triggered it again when trying to build icu4c statically to impress a friend of mine who does packaging for alpine. So until we have a handle on the infinite looping (#51245 narrows it down to variant defaults), I think this will be especially important.

Even once we do fix the infinite looping, this is the one part of the spack run that is completely unresponsive to user input or progress output.

Solution: push the solver to asyncio

• create spack.llnl.util.subthread to characterize threaded processes which produce output
• decouple polling for output from executing the process
• translate a repeated blocking poll into the language of asyncio using a thread pool executor

Result

• ^C works during solver hangs
• debug logs to show a sign of life even during infinite looping
• no performance loss