Fix bug that under-allocated compute nodes

[daniel-thom]

The scheduling algorithm was not correctly accounting for walltime in this case and scheduled half the required number of nodes.

[Copilot]

Pull request overview

Fixes an under-allocation bug in the Slurm scheduler planning logic where sequential batch capacity (“time slots”) was computed using the partition max walltime rather than the actual walltime that would be requested for the allocation under the selected walltime strategy.

Changes:

• Update allocation counting to compute time_slots from the actual allocation walltime (derived from WalltimeStrategy), not partition.max_walltime_secs.
• Update/expand HPC scheduler-generation tests to match the corrected allocation math and add a regression test for long-running whole-node jobs.
• Adjust the auto-merge CLI test to explicitly use --walltime-strategy max-partition-time so it continues to validate runtime-aware merging behavior.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
src/client/scheduler_plan.rs	Uses computed allocation walltime for time_slots so allocation counts reflect the requested walltime strategy.
tests/test_hpc.rs	Updates expected allocation counts and adds a regression test covering the previously under-allocating long-runtime case.
tests/test_slurm_commands.rs	Makes the test’s walltime assumptions explicit via --walltime-strategy max-partition-time.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.