Fix Normalize ignoring a single explicit in_min/in_max bound

[eminesu]

Summary

Normalize ignores an explicitly supplied in_min when in_max is None, and vice versa — the explicit value is silently discarded and both bounds are computed from the data percentiles instead. This happens because make_params only uses the explicit bounds when both are provided.

This PR resolves each input bound independently: an explicitly supplied bound is always honored, and only the missing one is computed from the percentiles. Behaviour when both bounds are given, or when both are None, is unchanged.

Reproduction (before this PR)

import torch
import torchio as tio

image = tio.ScalarImage(torch.tensor([0., 50., 100.]).reshape(1, 1, 1, 3))
print(tio.Normalize(out_min=0., out_max=1., in_min=50.)(image).data.flatten().tolist())
# before: [0.0, 0.5, 1.0]   (in_min=50 ignored)
# after:  [0.0, 0.0, 1.0]

Changes

• Normalize.make_params: resolve in_min / in_max independently and override the auto-computed bound(s) with any explicit value.
• Add regression tests for the in_min-only and in_max-only cases.

Fixes #1509