fix(solve): wait and collect results when --timeout 0

[lewisjared]

Description

ref solve --timeout 0 ran every diagnostic successfully into the scratch directory but copied nothing to the results directory and ingested nothing into the database.

Root cause

The default LocalExecutor only submits work in run(); all result handling (copying scratch → results, DB ingestion, mark_successful) happens inside join(). The solver only called join() when timeout > 0:

if timeout > 0:
    executor.join(timeout=timeout)

The CLI overloaded timeout as the "don't wait" sentinel (timeout=timeout if wait else 0), so --no-wait and --timeout 0 were the same thing internally. Passing --timeout 0 — which intuitively reads as "no time limit" — therefore skipped join() entirely, orphaning the completed work in scratch and leaving every Execution row stuck at successful=None.

Fix

• Decouple whether to wait from how long: solve_required_executions now takes an explicit wait and joins whenever wait is True, regardless of the timeout value.
• A non-positive timeout (<= 0) now means wait with no time limit in LocalExecutor.join and CeleryExecutor.join (HPC already did this via walltime).
• The CLI passes wait and timeout through separately instead of collapsing --no-wait onto timeout=0.
• Warn when --no-wait is used with an executor that only persists results during join (local, HPC), since skipping the join silently orphans completed work. This is driven by a new collects_results_on_join marker on each executor (True for local/HPC; False for synchronous, which collects in run(), and Celery, which collects via the handle_result worker callback).

Recovery for affected users

Existing scratch outputs from a previous --timeout 0 run can be recovered without re-computing via ref executions reingest --include-failed.

Checklist

Please confirm that this pull request has done the following:

• Tests added
• Documentation added (where applicable)
• Changelog item added to changelog/

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Flag	Coverage Δ
core	93.31% <100.00%> (+0.02%)	⬆️
providers	91.79% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...mate-ref-celery/src/climate_ref_celery/executor.py	97.61% <100.00%> (+0.05%)	⬆️
.../climate-ref-core/src/climate_ref_core/executor.py	86.27% <100.00%> (+0.56%)	⬆️
packages/climate-ref/src/climate_ref/cli/solve.py	100.00% <ø> (ø)
...ckages/climate-ref/src/climate_ref/executor/hpc.py	64.54% <100.00%> (+0.14%)	⬆️
...ages/climate-ref/src/climate_ref/executor/local.py	96.55% <100.00%> (+0.04%)	⬆️
...limate-ref/src/climate_ref/executor/synchronous.py	100.00% <100.00%> (ø)
packages/climate-ref/src/climate_ref/solver.py	98.90% <100.00%> (+0.37%)	⬆️

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.