Add an optional Numba-CUDA-MLIR Gaussian xFit backend

[tpn]

Gaussian fits repeatedly evaluate the model and analytic derivatives through CuPy expressions. Add an optional Numba-CUDA-MLIR backend that combines this work into separate residual and weighted-Jacobian CUDA kernels, with per-thread kernel reuse and explicit stream/buffer lifetimes. CuPy normal-equation contractions, the LM solver and final diagnostics are preserved.

Install cuphoton[numba-cuda-mlir] on Linux with CUDA 13 and select --backend numba-cuda-mlir. The selection also flows through device-pipeline configuration and bulk execution. Gaussian difference/split modes support float32/float64, masks and variance weights; sampled models and finite differences fail explicitly. Existing backend defaults and fallbacks are preserved.

Public synthetic tests cover scientific parity, solver edge cases, concurrent fits, delayed producer streams and actual CLI selection. The xScan-to-xFit inference test compares both backends through the CLI. GPU CI requires these tests to run without skips, and the existing Gaussian benchmark accepts the new backend for warmed comparisons. Builds on #82.

Validation on the integrated source (Python 3.12, CUDA 13):

• 707 CPU tests passed, covering xFit, shared execution, FITS controls, and pipeline contracts.
• Seven GPU CLI and end-to-end pipeline checks passed without skips, including difference/split fits and comparison with CuPy.
• All-file hooks passed with the optional GPU and Numba dependencies installed, including mypy and lock validation.

[coderabbitai]

Important

Review skipped

Auto reviews are limited based on label configuration.

🏷️ Required labels (at least one): 1

• ai-review

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

• Configuration used: Repository: NVIDIA/cuPhoton/.coderabbit.yaml
• Review profile: CHILL
• Plan: Enterprise
• Run ID: f4d7291d-6a12-4d90-911c-aa0434b1ac69

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}