Add script for scan-based summary backfill

[FestiveKyle]

The summaries job only recently started producing the psd/pgs/all scopes so there are years of missing history. This adds a one-off backfill that reconstructs those summaries for every past day.

• New scan_summaries.py, run as a K8s job, rebuilds chart + org summaries directly from the raw dns/web scans (not the denormalized domain.status), so history can be reconstructed for any date.
• Writes to shadow collections chartSummaries_rebuild / organizationSummaries_rebuild, prod collections are untouched. (--target prod exists but isn't used here.)
• Carries each domain's status forward per scan-type, matching the live job's domain.status semantics.

After ensuring the new summaries data aligns with the previously generated data, we can switch to use --target prod to move to the prod collections. At that point we should also merge the logic of scan_summaries.py with summaries.py.