Parse SchedMD pre-release / zero-padded Slurm versions robustly

[sdmcclain]

Summary

slurm_version() parsed the Slurm version with tuple(int(v) for v in version.split(".")), which assumes every dot-separated component is a plain integer. SchedMD pre-release builds append a -<n>pre<n> suffix to the micro version, so sinfo -V reports e.g. slurm 26.05.2-0pre1 and int("2-0pre1") raises ValueError: invalid literal for int() with base 10: '2-0pre1'. This crashes every caller — for example gcm's slurm_monitor poller — on clusters running Slurm built from a SchedMD development branch rather than a tagged release.

Parse the numeric release with packaging.version.Version after stripping the non-PEP440 pre-release suffix. packaging also normalizes the zero-padded YY.MM minor (26.05 -> 26), which strict SemVer rejects outright. The Tuple[int, ...] return type is unchanged so existing comparisons (e.g. slurm_version() >= (23, 2)) keep working, and any unexpected string falls back to the original split. Adds packaging (pure-Python, already ubiquitous) to dependencies.

Test Plan

Exercised the patched slurm_version() with the version string mocked to each SchedMD form — including the zero-padded pre-release the old int() split crashed on:

$ python -c "
from unittest.mock import patch
import clusterscope.lib as lib
for s in ['26.05.2-0pre1', '24.11.2-0pre1', '24.05.0', '25.11']:
    with patch.object(lib, 'get_unified_info') as m:
        m.return_value.get_slurm_version.return_value = s
        print(f'{s:16} -> {lib.slurm_version()}')
"
26.05.2-0pre1    -> (26, 5, 2)
24.11.2-0pre1    -> (24, 11, 2)
24.05.0          -> (24, 5, 0)
25.11            -> (25, 11)