General bug fixes and deprecation updates

[CarterDW]

Updated numpy, scipy, matplotlip requirements, removed deprecated np.trapz as well as np.rowstack for np.trapezoid and np.vstack, as well as updated .toml, readme, and .yml to reflect new changes

[Copilot]

Pull request overview

This PR modernizes the project’s Python/scientific-stack baseline by raising minimum dependency versions (NumPy/SciPy/Matplotlib), bumping the minimum supported Python version, and updating code to avoid NumPy APIs that are deprecated/removed in newer NumPy releases.

Changes:

• Raised minimum versions for numpy, scipy, and matplotlib, and increased requires-python to >=3.9.
• Replaced deprecated NumPy APIs (np.trapz, np.row_stack) with supported equivalents (np.trapezoid, np.vstack) across the codebase.
• Updated examples and CI configuration to reflect the new baseline (with a noted CI version issue).

Reviewed changes

Copilot reviewed 22 out of 23 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
requirements.txt	Updates minimum dependency versions for end users installing via requirements file.
README.md	Removes outdated note about old NumPy compatibility.
pyproject.toml	Updates requires-python and project dependencies to match new minimum versions.
fastpt/utils/OV.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/utils/matter_power_spt.py	Removes legacy/commented trapz import reference.
fastpt/utils/kPol.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/utils/J_table.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/rsd/RSD.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/IA/IA_tt.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/IA/IA_ta.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/IA/IA_gb2.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/IA/IA_ctbias.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/IA/IA_ct.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/IA/IA_ABD.py	Replaces np.row_stack with np.vstack for NumPy 2.x compatibility.
fastpt/core/FPTHandler.py	Fixes docstring to reference save_instance() as the producer for load_instance().
fastpt/core/FASTPT.py	Replaces np.trapz with np.trapezoid in sig4 computation.
fastpt/core/FASTPT_simple.py	Replaces np.trapz with np.trapezoid in bias integral computation.
fast_pt.egg-info/requires.txt	Updates generated metadata dependency versions (should not be version-controlled).
examples/IA_ABD_plot.py	Clarifies LaTeX dependency for usetex=True.
examples/hello_fastpt.py	Updates example to use save_instance() API.
examples/fastpt_example.py	Updates example to retrieve sig4 from one_loop_dd_bias return tuple.
examples/fastpt_example_nonlin_bias.py	Changes plotting behavior (now interactive), impacting headless execution.
.github/workflows/ci.yml	Updates Actions versions and dependency install steps; adds Python 3.14 (problematic).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[CarterDW]

@copilot re review

[Copilot]

@copilot re review

I re-reviewed the current branch. I didn’t find any additional actionable changes from the prior review threads, so no new code changes were needed in this pass.

[athoyas]

These skipif decorators skip the benchmark comparisons on ARM64, which means those terms (IA_ta, IA_ct, IA_mix, the RSD ones) have no regression check at all on Apple Silicon. I compared the regenerated baselines against master — the drift is small (~1e-5 absolute, just NumPy 1.x→2.x floating-point differences), so could we loosen the tolerance instead of skipping? Something like np.allclose(..., rtol=1e-3, atol=1e-4) would pass on all platforms while still catching real breakage.
Also, the skip conditions look off: sys.version_info <= (3, 14) is true for basically every current Python, so combined with or arm64 these skip almost everywhere. And a couple of the reason strings say "3.13" while the check is <= (3, 14).

[athoyas]

While tracing the sig4 indexing change through one_loop_dd_bias, I noticed something unrelated in _get_Ps2s2 (in FASTPT.py). Not caused by this PR, but might be worth addressing. The function computes and returns its result in a local variable named Pd2s2 rather than Ps2s2. Looks like a copy-paste leftover from _get_Pd2s2 directly above it. Functionally it seems harmless — the value is cached under the correct "Ps2s2" label and returned normally, so callers get the right result. Worth renaming for clarity (I didn't verify the formula itself).

[CarterDW]

Differs from marcpaterno PR by also updating np.rowstack and fixing a few pytest failings and other various bookkeeping issues