Guard get_shares_full against empty/missing share-count timeseries

[joaopedroassad]

Summary

• TickerBase.get_shares_full reads json_data["timeseries"]["result"] and then indexes shares_data[0] with no guard. When Yahoo returns a 200 whose share-count timeseries envelope has an empty result ([]), a null result, or omits timeseries entirely (the finance error envelope), this raises a raw IndexError / TypeError / KeyError that leaks straight to the caller and ignores hide_exceptions.
• Empty result is the common case: symbols with no share-count history (a lot of non-US tickers, ETFs, indices) come back as {"timeseries": {"error": null, "result": []}}, so shares_data[0] hits IndexError: list index out of range.
• The break isn't limited to the public get_shares_full() call. Quote.shares calls get_shares_full internally to populate the shares complementary field, so the same exception can take down .info for the whole ticker.
• Fix is minimal and matches how the rest of the function already degrades: pull result defensively and return None when it's empty, exactly like the existing if "shares_out" not in shares_data[0]: return None exit right below it. Every other failure path in this function already returns None.

shares_data = (json_data.get("timeseries") or {}).get("result") or []
if not shares_data or "shares_out" not in shares_data[0]:
    return None

This is the same class of empty/malformed-timeseries guard as #2877 (Complementary info: guard against empty or missing timeseries result), which fixed the identical json_result["result"][0] pattern in quote.py. This is the next-door site in base.py that reads the same share-count timeseries shape and was still unguarded.

Testing

Ran in a fresh venv with an editable install:

python -m venv .venv && . .venv/bin/activate
pip install -e . pytest ruff

Red -> green on the new test:

# stock dev (fix reverted): FAILS with IndexError: list index out of range at base.py:541
python -m pytest tests/test_ticker.py::TestTickerInfo::test_shares_full_empty_result_timeseries -v
# with the fix: PASSES
python -m pytest tests/test_ticker.py::TestTickerInfo::test_shares_full_empty_result_timeseries -v

Adjacent complementary-info guard tests still pass:

python -m pytest tests/test_ticker.py::TestTickerInfo::test_complementary_info_sparse_timeseries tests/test_ticker.py::TestTickerInfo::test_complementary_info_empty_result_timeseries -v

Lint / compile clean on changed files:

ruff check yfinance/base.py tests/test_ticker.py    # All checks passed!
python -m py_compile yfinance/base.py tests/test_ticker.py
git diff --check

The new test is fully mock-based (patches _data.cache_get, sets _tz to skip the network tz lookup), so it needs no network. It covers the four failing shapes (empty list, missing result, null result, missing timeseries) plus a happy path that still returns a populated Series.

[ValueRaider]

When Yahoo returns a 200 whose share-count timeseries envelope has an empty result

Example?