Add validation for detector module/gap geometry inference

[coderabbitai]

Context

During review of PR #45 (#45 (comment)), a potential issue was identified in mdx2/dxtbx_machinery.py where detector module and gap calculations could silently produce incorrect geometry.

Requested by: @spmeisburger
Related PR: #45
File: mdx2/dxtbx_machinery.py (lines ~21-75)

Problem

The current code computes num_modules_* and gap_* using integer division (//), which can silently hide remainders and produce plausible-but-wrong geometry values. This could lead to incorrect downstream binning/tiling without any error indication.

Requested Solution

Phase 1: Unit Tests (Required First)

Create comprehensive unit tests for the detector geometry validation in Experiment.__init__:

• Test cases with valid, evenly-divisible detector layouts
• Test cases with invalid layouts (non-divisible gaps, wrong module sizes)
• Edge cases (single module, zero gaps, etc.)
• Boundary conditions

Phase 2: Implementation

Add explicit validation checks to ensure:

1. Module sizes (self._module_size_fast and self._module_size_slow) are > 0
2. num_modules_fast and num_modules_slow (computed as image_size[axis] // module_size) are >= 1
3. total_gap_* (computed as image_size[axis] - (num_modules_* * module_size)) is >= 0
4. When num_modules_* > 1, verify that total_gap_* % (num_modules_* - 1) == 0 (gaps must be evenly divisible)
5. Raise clear RuntimeError messages when validation fails

This will make geometry failures explicit instead of producing silently wrong values.

Acceptance Criteria

• Unit tests created covering valid and invalid detector geometries
• All tests initially demonstrate current behavior
• Validation logic added to Experiment.__init__
• Tests pass with new validation
• Clear error messages guide users when geometry is invalid

References

• Note: This issue replaces Add validation for detector module/gap geometry inference ando-lab/mdx2#9 (created in wrong repository)