feat: widen the prepared parameter types (BIGINT, DOUBLE, NUMERIC, temporal, binary) #840

Problem

The prepared encoders (_encode_prepared_scalar, used by pycubrid.compat.native bind_param() and by any prepared batch path) accept only a 32-bit INT, a string and NULL, plus INT/STRING collections and LOB handles. In the #836 probes they rejected bool, an int beyond 32 bits, float, Decimal, date, datetime and bytes. The literal executemany() path renders all of these. Without the wider set, an opt-in EXECUTE_ARRAY path (#839) can serve only INT/string workloads.

Scope

Add typed prepared encodings with offline tests, plus integration round trips on CUBRID 10.2-11.4:

• BIGINT for 64-bit ints;
• DOUBLE for float, rejecting nan and inf as the literal path does;
• NUMERIC as text for Decimal;
• DATE, TIME and DATETIME (and DATETIMETZ if the server accepts it) for the temporal types;
• VARBIT for bytes;
• a documented choice for bool.

Document each mapping next to the literal mapping in PARAMETER_BINDING.md. Note where host-variable coercion differs from literal coercion.

Non-goals

• Changing the literal binding contract of ordinary cursors.

Links

• chore(performance): evaluate a prepared-statement batch path for executemany #836 (decision), experiment: pycubrid prepared batch paths, including EXECUTE_ARRAY, on CUBRID 10.2-11.4 cubrid-benchmark#23 (probe data: semantics_<version>.json)