fix: surface valid table names on SQL validation 'table not found' retry

[ghostiee-11]

When the SQL agent generates a query that references a table that doesn't exist (e.g. the source name instead of one of its tables, which the model does for multi-table sources like XArraySQLSource where each variable is its own table), validation fails with "table not found". The retry feedback was just the raw error string, so the model kept making the same mistake and exhausted its retries (RetriesExceededError). This surfaces the source's actual table names in the retry feedback so the model recovers.

Before

table 'UploadedSource000002' not found
  -> retry feedback = raw error -> model guesses again -> RetriesExceededError

After

table 'UploadedSource000002' not found
  -> retry feedback: "Use one of these exact table names, not the source name:
     'CMI_C01', 'DQF_C01', ..."
  -> model selects CMI_C01 -> validates

Surfaced while testing the Arraylake source control (#1889) against a real multi-table GOES XArraySQLSource. Adds two deterministic tests for _validate_sql (no LLM needed).

[codecov]

Codecov Report

❌ Patch coverage is 95.55556% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 70.66%. Comparing base (7cc94fb) to head (d8b5609).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
lumen/tests/ai/test_sql_validation.py	95.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1890      +/-   ##
==========================================
+ Coverage   70.60%   70.66%   +0.05%     
==========================================
  Files         195      196       +1     
  Lines       33630    33675      +45     
==========================================
+ Hits        23746    23795      +49     
+ Misses       9884     9880       -4     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ahuang11]

What happens if there are over 100 of tables in a database? Won't that consume too much context? I think we may need to do similarity search, perhaps through Metaset?

[ghostiee-11]

Good catch on the ordering. It's already capped at [:50] so context stays bounded, but get_tables() comes back in arbitrary order, so with a lot of tables the slice can drop the one the model actually needs.

The metaset's a better fit, and it's already in context here. get_top_tables() ranks by similarity to the query, so I'll pull the relevant names for that source from it and only fall back to get_tables() when it's empty. Pushing that shortly.

[ahuang11]

I think we need to mention it's truncated if it's over 50