fix: raise `ValidationError` for decimal precision outside range (#3585)

Closes #3583

# Rationale for this change
According to the Iceberg specification, decimal precision must be
between 1 and 38. Previously, `pyiceberg` silently accepted invalid
values (such as `precision=39` or `0`), which could lead to data
corruption or crashes downstream when encoding/decoding fixed-byte
decimals (matching the Java implementation's
`IllegalArgumentException`).

This PR adds a defensive validation guard into `DecimalType.__init__` to
raise a `ValidationError` when the precision is out of bounds `[1, 38]`.

Additionally, fixed an existing invalid type definition
(`DecimalType(100, 2)`) inside `tests/test_schema.py` which was
violating the specification bounds and causing the new validation guard
to trip during the integration test suite.

## Are these changes tested?
Yes, added dedicated unit tests in `tests/test_types.py`:
- `test_decimal_precision_validation` covers upper bounds
(`precision=39`), lower bounds (`precision=0`, `precision=-5`), and
ensures valid boundary values (`precision=38`) work perfectly.

Also updated existing test assertions in `tests/test_schema.py` to use a
compliant precision value (`38`).

## Are there any user-facing changes?
Yes, creating a `DecimalType` with an invalid precision will now
explicitly raise a `pyiceberg.exceptions.ValidationError` instead of
failing silently at runtime.