perf: replace KeyError exception with dict.get() in BoundStatement.bind()

[mykaul]

Summary

• Replace try/except KeyError with dict.get() + sentinel in the per-column binding loop of BoundStatement.bind()
• This loop runs once per column per session.execute() call for dict-style bindings, making it a hot path
• The sentinel _BIND_SENTINEL = object() distinguishes missing keys from explicit None values

Motivation

When binding a dict to a prepared statement, the driver iterates over every column in the prepared statement metadata and looks up the value in the user-supplied dict. The previous code used try: values_dict[col.name] / except KeyError: which raises and catches a KeyError for every missing/optional column. While CPython 3.11+ made try/except cheaper when no exception fires, the exception does fire for every omitted column, and raising exceptions is inherently expensive (~5-10x slower than a dict.get() + identity check).

For workloads using protocol v4+ (the common case), omitted columns are bound as UNSET_VALUE, so the KeyError path fires frequently for partial-column updates.

Changes

• cassandra/query.py: Add module-level _BIND_SENTINEL = object() sentinel, replace the try/except block with dict.get(col.name, _BIND_SENTINEL) + is not check

Testing

• All existing unit tests pass (tests/unit/test_query.py, tests/unit/test_cluster.py, tests/unit/test_resultset.py)
• Semantic behavior is unchanged: same values bound, same errors raised for pre-v4 protocol