Fix parsing of Polygon statement metadata

[mfornet]

This is a patch which I'm using locally due to some newer errors. Everything was written by codex.

---

Summary

• ignore the encoding entry when turning problem.statements fields into local statement files
• ignore non-string statement metadata fields returned by newer Polygon API responses
• add regression coverage for current metadata and non-UTF-8 statement encodings

Bug

ProblemSession.get_statements_list() currently assumes that every value in each language returned by problem.statements is textual statement content.

Newer Polygon responses can also contain boolean metadata such as:

showCautionsAndGrammaticalFixes: false
showInReview: true

The method passes those booleans to PolygonFile.to_byte(), which calls .encode() and crashes:

AttributeError: 'bool' object has no attribute 'encode'

As a result, commands that enumerate remote files, including polygon-cli update and polygon-cli list, fail for affected problems. The existing loop also incorrectly treats the string-valued encoding metadata as a downloadable encoding.tex statement section.

Fix

Skip encoding and any non-string values before constructing PolygonFile statement entries. Actual statement sections keep using the language's declared encoding, preserving the existing behavior for legacy API responses.

Validation

• python3 -m unittest discover -s tests -p 'test_*.py' -v
• live polygon-cli update against a problem containing English and Spanish statements plus the new boolean metadata; the update completed and every tracked file reported Not changed