Fix epoch gaincal calibrator tile selection

[jakobtfaber]

Summary

• rank every hourly-epoch MS by the brightest matched VLA calibrator, using midpoint separation as the same-calibrator tiebreaker
• fall back to bright-source counts across all tiles instead of only the center pair
• continue from a missing VLASS strip database to NVSS source counts

Validation

• .......................... [100%]
  =============================== warnings summary ===============================
  :488
  :488
  :488
  :488: DeprecationWarning: builtin type swigvarlink has no module attribute

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
26 passed, 3 warnings in 4.34s — 26 passed

• live probes select , , and
• no new Ruff findings relative to ; existing file-level findings remain unchanged

Bounded H17 confirmation

The Jul 15 calibrator tile raises median unflagged p.G SNR to 57.09, but the production preconditioner propagates a 46.15% flag mask, so the 30% gate still fails. A no-preconditioner control improves to 33.33% flagged / SNR 65.70 but also remains above the gate. This PR fixes the confirmed selection/catalog defects without changing or the gate threshold; full Jul 12/13/15 reruns remain blocked pending the residual antenna/baseline-coherence investigation.