Normalize raw session metadata before ingest in the nightly cron

[lecriste]

What

Add a step to the nightly local cron that reconciles the per-session index metadata stored in the raw acquisition files (.npy) with the database before
populate_rig ingests them.

• docker/cron_common.sh: new cron helper that runs the repo's existing raw-metadata resync entrypoint in the client container, using the same setup as the scenario runner (vr4mice_check_env_file + BASE_INSTALL + source .env).
• cron_script.sh: invoke it after compose up and before vr4mice_run_cron_scenario local.

The resync action itself already exists in the repo; this PR only invokes it from the cron.

Why

Some session tables (exp.Session / SessionScoreSheet) use a per-session index that is baked into the raw files as part of their primary key. A duplicated value colliding with an earlier session makes the new session silently skipped on ingest — it never lands in the database, with no error. The resync recomputes the
correct values from the database, removing the collision at the source before ingest.

Behavior / notes

• Runs in the client container against /data/data, after compose up (containers up) and before populate_rig moves files to processed/.
• Fatal on failure: cron_script.sh uses set -euo pipefail, so if the resync (or its env check) fails, the run aborts and the raw files are left untouched in data/data for the next run — rather than ingesting inconsistent metadata.
• Idempotent: the resync rewrites a file only when the stored value differs from the computed one, so already-correct files are a no-op.
• Local job only: added to cron_script.sh and not cron_script_aws.sh.
• No crontab change — the step lives inside the script.

Testing

Ran the resync entrypoint on the production server: it corrected freshly-acquired files whose stored index was inconsistent, and was a verified no-op on already-correct files.
The new cron helper is a thin wrapper that runs exactly that entrypoint with the standard cron environment.