Description:

Python's native float type is a 64-bit IEEE 754 double-precision number. However, the dbsql_parameter_from_primitive() function in src/databricks/sql/parameters/native.py (line 662) maps it to FloatParameter, which casts the value to a Databricks SQL FLOAT (32-bit single-precision). This causes silent precision loss when using parameterized queries with float values.

The correct mapping should be DoubleParameter, which corresponds to Databricks SQL DOUBLE (64-bit) and matches Python's native float precision.

Steps to reproduce:

from databricks import sql
from databricks.sql.parameters import DoubleParameter

test_val = 12345.678901234

with connection.cursor() as cursor:
    cursor.execute("DROP TABLE IF EXISTS test_precision")
    cursor.execute("CREATE TABLE test_precision (val DOUBLE)")

    # BUG: raw float silently loses precision (uses FloatParameter → 32-bit)
    cursor.execute("INSERT INTO test_precision (val) VALUES (?)", [test_val])
    cursor.execute("SELECT val FROM test_precision")
    stored_bug = cursor.fetchone()[0]

    print(f"Input:  {test_val}")       # 12345.678901234
    print(f"Stored: {stored_bug}")     # 12345.6787109375 (precision lost!)
    print(f"Match:  {test_val == stored_bug}")  # False

Expected Behavior

Passing a Python float to a parameterized query should preserve 64-bit precision by default, since Python's float is inherently 64-bit.

Actual Behavior

Precision is silently truncated to 32-bit because dbsql_parameter_from_primitive() returns FloatParameter instead of DoubleParameter.