Update CHANGELOG.md for v1.0.2 (unreleased) and clarify changelog policy

[dweindl]

Summary

• Adds CHANGELOG.md entries for everything merged to main since v1.0.1 that wasn't yet documented, based on the actual PR descriptions:
  • BREAKING CHANGES: Python >=3.12 requirement (Require Python>=3.12 #3180)
  • Features: JAX adjoint checkpoints parameter (Add checkpoints parameter for reverse-mode adjoint control #3217); PEtab SciML equinox module import (Add equinox module import support for petab sciml problems #3179), new layer types (Implement missing layer types  #3176), PEtab v2 support (Update PEtab SciML support to PEtab v2 #3165)
  • Fixes: model-import AttributeError (Fix AttributeError during model import #3200), PEtab flattening noise-formula bug (Fix noise formula substitution in flattening of PEtab problem #3177), conservation-law assertion fix (Fix spurious conservation-law assertion for condition-dependent compartments #3202), and 5 JAX/PEtab-backend fixes (Fix NameError in JAX _known_discs for w-dependent explicit triggers #3195, Fix condition-name-based preequilibration detection in JAX PEtab backend #3205, Fix JAXProblem.save()/load() round-trip against current petab.v2 API #3204, Fix JAX PEtab condition-table resolution: state reinit, noise distribution, and differentiability #3207, Fix JAX PEtab backend: discontinuities, PEtab v1 import, event assignments #3196, Fix JAX benchmark failures at the root cause (unblock Benchmark Collection JAX) #3210)
  • Performance (new section): JAX compile-time fix (perf(jax): cut JAX PEtab benchmark compile time via CSE + ImplicitAdjoint #3212), JAX recompilation-avoidance fix (Avoid JAX recompilation from fresh steady_state_event closures #3197), and the two CMake configure speedups (Skip redundant HDF5 detection during model CMake configure #3220, Skip redundant compiler-flag checks in CMake configure #3221)
  • Documentation (new section): PEtab SciML example notebook (Add PEtab SciML example notebook #3208)
• Pure CI/dependency-bump commits and internal test-only changes were deliberately left out — none of those are user-facing.
• Also clarifies the changelog-entry guidance in doc/development.rst: it covers any user- or contributor-visible change, including performance/build-system changes that affect model import or build times, not just new APIs — and should be omitted when the change has no user-visible effect (typos, CI-only, private refactoring).

Test plan

• Reviewed each referenced PR's actual description (not just commit titles) to write accurate entries
• Cross-checked the commit list between v1.0.1 and main to confirm no other user-facing changes were missed

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77.96%. Comparing base (eb22b44) to head (959cebf).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3222      +/-   ##
==========================================
- Coverage   78.56%   77.96%   -0.61%     
==========================================
  Files         318      318              
  Lines       21090    21090              
  Branches     1487     1487              
==========================================
- Hits        16570    16442     -128     
- Misses       4512     4640     +128     
  Partials        8        8              

Flag	Coverage Δ
cpp	71.97% <ø> (ø)
cpp_python	36.59% <ø> (ø)
petab	48.27% <ø> (ø)
petab_sciml	16.22% <ø> (ø)
petab_sciml_benchmarks	14.78% <ø> (ø)
python	70.34% <ø> (ø)
sbmlsuite-jax	?

Flags with carried forward coverage won't be shown. Click here to find out more.
see 4 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.