Add GPU option for time series estimation

[mohseniaref]

Summary

• implement optional GPU acceleration in ifgram_inversion.py
• add use_gpu parameter to estimate_timeseries and calc_inv_quality

Testing

• pytest -q
• python -m compileall -q src/mintpy

---

https://chatgpt.com/codex/tasks/task_e_6848ab8ab8748320baac958522fa6f13

[Copilot]

Pull Request Overview

This PR adds optional GPU acceleration to the time-series estimation workflow and provides a benchmarking utility.

• Introduce a use_gpu flag in estimate_timeseries and calc_inv_quality to switch between NumPy and CuPy backends
• Adapt core linear algebra and array operations to use xp (NumPy or CuPy)
• Add benchmark_gpu_speedup to compare CPU vs. GPU performance

Comments suppressed due to low confidence (3)

src/mintpy/ifgram_inversion.py:146

• [nitpick] The variable name linmod is not very descriptive. Consider renaming it to something like xp_linalg or backend_linalg for clarity.

linmod = linalg

src/mintpy/ifgram_inversion.py:93

• There are no tests covering the use_gpu=True path. It would be helpful to add unit tests that run with use_gpu=True to validate GPU behavior and prevent regressions.

print_msg=True, use_gpu=False):

src/mintpy/ifgram_inversion.py:418

• The time module isn't imported in this file, so calls to time.perf_counter() will raise a NameError. Please add import time at the top.

t0 = time.perf_counter()