fix(adapters): resolve DuckDB extension, Arrow NULL-typing, and MSSQL row-format bugs

[cofin]

Fixes three independent adapter bugs, each with regression coverage.

#631 — select_to_arrow all-NULL columns lose their type

pa.Table.from_pylist infers types from values only, so a column that is NULL in every row collapsed to Arrow null type and lost the DB column's declared type. Downstream consumers misread it (DuckDB materializes null as INTEGER, so COALESCE(col, 'x') raised a conversion error). convert_dict_to_arrow now retypes value-inferred null columns to string, which safely holds the all-null values.

Closes #631

#636 — DuckDB emits two warnings for one extension setup failure

A non-required extension whose explicit install failed and whose subsequent load also failed emitted both pool.extension.install.failed and pool.extension.load.failed for one unusable extension. Install + load are now treated as one setup outcome:

• install-fail + cached-load-success → no warning (install detail logged at DEBUG)
• install-fail + load-fail → one pool.extension.setup.failed carrying both install_error and load_error
• install-success + load-fail → one pool.extension.setup.failed
• name-only load-only and required=True contracts unchanged

Closes #636

#630 — MSSQL marks rows row_format="tuple" but driver returns mssql_python.Row

mssql_python.Row is iterable/indexable but not a tuple subclass, so declaring row_format="tuple" tripped the debug row-format fidelity assertion during dict materialization. dispatch_execute now materializes fetched rows into real tuples (new materialize_tuple_rows helper), matching the tuple row shape used across the SQL Server adapter family.

Closes #630

Testing

• New/updated unit tests for all three (arrow null-typing, DuckDB setup-outcome collapse, MSSQL Row→tuple materialization)
• make type-check clean (mypy + pyright)
• Verified downstream: DuckDB COALESCE on an all-NULL column, and SQLResult materialization of Row-like objects under row_format="tuple"

[codecov-commenter]

Codecov Report

❌ Patch coverage is 95.65217% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 76.82%. Comparing base (6bb5742) to head (e087c6e).
⚠️ Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
sqlspec/adapters/mssql_python/driver.py	50.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #641   +/-   ##
=======================================
  Coverage   76.82%   76.82%           
=======================================
  Files         473      473           
  Lines       66108    66123   +15     
  Branches     9821     9825    +4     
=======================================
+ Hits        50785    50799   +14     
+ Misses      12031    12029    -2     
- Partials     3292     3295    +3     

Flag	Coverage Δ
integration	60.01% <39.13%> (-0.03%)	⬇️
py3.10	75.08% <95.65%> (+0.01%)	⬆️
py3.11	75.10% <95.65%> (+0.01%)	⬆️
py3.12	75.09% <95.65%> (+<0.01%)	⬆️
py3.13	75.09% <95.65%> (+0.01%)	⬆️
py3.14	76.02% <95.65%> (+0.01%)	⬆️
unit	63.94% <95.65%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
sqlspec/adapters/duckdb/pool.py	79.48% <100.00%> (+0.44%)	⬆️
sqlspec/adapters/mssql_python/core.py	82.03% <100.00%> (+0.57%)	⬆️
sqlspec/utils/arrow_helpers.py	72.94% <100.00%> (+1.85%)	⬆️
sqlspec/adapters/mssql_python/driver.py	76.56% <50.00%> (ø)

... and 5 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.