Cache a progress report in dscheck.einfo before the PBS walltime

[zaihuaji]

Summary

• A long -A 3 / -A 4 batch run was killed by PBS before start_actions() reached its email block, so no report was ever sent. The run now caches a progress report into dscheck.einfo an hour before the walltime and keeps working; the dscheck daemon mails it once the record is unlocked (after a kill or a normal finish).
• A run that finishes in time overwrites the cached progress report with the full report, so exactly one mail goes out.
• The walltime is read back from PBS (qstat ReqdTime) instead of assumed, since the dscheck daemon caps the requested 24 h to the maximum of the queue the job lands in (only 6 h for default).
• Guide deck slides 20/23/16 updated to match; the old "stops cleanly at 23 hours" description no longer applied.

Test plan

• python3 -m py_compile src/rda_python_dsquasar/dsquasar.py
• Verified get_pbs_info() returns {} on failure and that the ReqdTime key is produced (pg_sig.py:919)
• Verified build_customized_email() sends first and only caches on failure, so cache_customized_email() is required to park a report (pg_dbi.py:2190/2277)
• Verified the daemon only acts on einfo when pid = 0 (pg_check.py:436/455), so a cached report is not mailed while the job still runs
• Verified both check_batch_deadline() callers run in the parent only (children sys.exit() at dsquasar.py ~922/~1014)
• Deck regenerated and QA'd from the exported PDF

Generated with Claude Code