TimeSeries: clarify required columns error messages (swev-id: astropy__astropy-13033)

[casey-brooks]

Summary

• refine BaseTimeSeries._check_required_columns to emit list-style errors when multi-column requirements are violated
• add regression coverage using TimeSeries to assert the clearer message for missing secondary required columns
• ensure single-column validations retain their previous messaging

Reproduction Steps

1. Create a TimeSeries and set ts._required_columns = ['time', 'flux']
2. Insert a flux column at index 1, then remove it so the table starts ['time', 'a']
3. Trigger the automatic check via ts.remove_column('flux')

Observed Failure (before this change)

ValueError: TimeSeries object is invalid - expected 'time' as the first column but found 'time'

New Behavior

• Multi-column requirements now raise
  • ... required ['time', 'flux'] as the first columns but time series has no columns, or
  • ... required ['time', 'flux'] as the first columns but found ['time', 'a']
• Single-column error messages are unchanged

Fixes #65

Testing

• LD_LIBRARY_PATH=$HOME/.nix-profile/lib .venv/bin/pytest tmp_test_required_columns.py

[casey-brooks]

Local verification:\n- LD_LIBRARY_PATH=/root/.nix-profile/lib .venv/bin/pytest tmp_test_required_columns.py (tests: 1 passed, 0 failed, 0 skipped)\n- LD_LIBRARY_PATH=/root/.nix-profile/lib .venv/bin/flake8 astropy/timeseries/core.py astropy/timeseries/tests/test_sampled.py (lint: no issues)

[noa-lucent]

Verified that BaseTimeSeries._check_required_columns preserves the single-column error message while multi-column prefixes report required/found lists. The new sampled time series test asserts the exact string when removing a required non-time column, and existing coverage for single-column behavior remains intact. Style touch-ups stay within project conventions.