Fix beam pointing antenna-health fallback

[peijin94]

Summary

• Validate antenna-health payloads read from DSA Store and raise clear errors for missing, incomplete, or inconsistent
  data.
• Fall back from the caltable bad-antenna list to selfcorr when caltable state is unavailable.
• Normalize both LWA-002A and 002A antenna-name formats before converting them to correlator numbers.
• Add regression tests for the fallback and both antenna-name formats.

Why

Beam 5 scheduled recordings were starting, but each scheduled pointing update aborted while reading /mon/anthealth/ caltable. The missing value was dereferenced as a dictionary, producing:

'NoneType' object is not subscriptable

Consequently, no fresh beam coefficients were sent.

Using selfcorr directly exposed a second issue: its antenna names already include the LWA- prefix, while
control_bf always added another prefix. Names such as LWA-002A therefore became LWA-LWA-002 and failed
correlator mapping.

This patch lets pointing continue with the available selfcorr health list and handles either supported antenna-name
representation.

Validation

• 9 focused unit tests pass under the deployed deployment Python environment.
• python -m py_compile mnc/control.py mnc/anthealth.py tests/test_control.py passes.
• The live beam 5 scheduler resumed periodic Sun tracking.
• Fresh beam coefficients were issued on all 32 pipelines.
• The beam 5 UDP power stream remained finite and nonzero after pointing recovery.

Modified files:

• mnc/control.py
• mnc/anthealth.py
• tests/test_control.py

Deployment safety and rollback

Before applying the live recovery patch, copies of the original deployed files were preserved on lwacalim10:

• /opt/devel/pipeline/envs/deployment/lib/python3.6/site-packages/mnc/control.py.pre-beamfix-20260817
• /opt/devel/pipeline/envs/deployment/lib/python3.6/site-packages/mnc/anthealth.py.pre-beamfix-20260817

These backups provide a direct rollback path if the deployed changes need to be reverted. The scheduler was not
restarted during the active observation.

[caseyjlaw]

I think a patch to improve reliability is good. However, it sounds like the /mon/anthealth/caltable etcd key is not being filled. There is a function caltable_flags that is used to fill that, but I think it is run manually when new tables are available.
Could you test that function with your change on your tables? I would like to be sure that the antenna naming syntax is consistent with your change.

[caseyjlaw]

@peijin94 I'm sorry, but you'll need to rebase after my latest commit to main.