[Python] OverflowError when inferring large unsigned integers (uint64 not handled)

### Describe the bug, including details regarding any error messages, version, and platform.

When converting Python integers to Arrow types, the current type inference
always defaults to int64. This causes an OverflowError when handling integers
greater than 2**63 - 1.

Example:
data = [0, np.iinfo(np.uint64).max]
pa.array(data) # raises OverflowError

Expected:
Inference should detect values that fit into uint64 and select that type
instead of failing.

A proposed fix tracks min/max integer values during inference and chooses the
smallest fitting signed/unsigned integer type.

pr link: https://github.com/apache/arrow/pull/47604

### Component(s)

Python