Prevent blank CSV fields in generated AMY EPW files

[ericmartinpe]

Summary

create_amy_epw_file could write blank fields into the output EPW file when AMY values were missing and not fully resolved by interpolation/imputation, producing malformed CSV rows (double commas). This PR ensures every observation field is always populated, and fixes a related issue where the output directory wasn't created if it didn't already exist.

Changes

• create_amy_epw_file.py:
  • Auto-create amy_epw_dir via os.makedirs if it doesn't already exist, instead of failing.
  • Forward/backward-fill the TMY Flags column so synthesized leap-day rows don't leave the EPW flag field empty.
  • For Tdb, Tdew, Patm, Wdir, and Wspeed: any AMY value still missing after interpolation/imputation now falls back first to the corresponding TMY observation, then to forward/backward fill, and finally to the EPW-standard missing-value sentinel (e.g. 99.9, 999999, 999) as a last resort — so no field is ever left blank.
• tests/test_create_amy_epw_file.py:
  • Added test_no_empty_fields_in_observation_rows, which generates a file and asserts no observation row contains a ,, (empty field).
  • Added test_creates_missing_output_directory, which verifies amy_epw_dir is created automatically (including nested paths) and the resulting file is valid.

Test plan

• New unit tests added and pass
• Full existing test suite run to confirm no regressions (21/21 passed)