DZTR 0.3.1: repository and packaging cleanup

DZTR 0.3.1 packages the repository and distribution cleanup following 0.3.0.

• Removes 11 notebook checkpoint copies and six Python bytecode files from the
  tagged source. Future release checks reject tracked generated files.
• Excludes generated caches, notebook checkpoints and AGENTS.md from the
  application source ZIP. AGENTS.md stays in the repository for development.
• Replaces personal workflow wording and machine-specific commands in public
  documentation with neutral researcher instructions.
• Selects release metadata from the current distribution version, including
  its release date, tag and DOI, rather than the previous version's metadata.

The reservoir equations, observation handling, parameterization methods,
optimization strategies, calibrated parameters and frozen Gardiner, Oahe and
Garrison examples are unchanged. No new production calibration jobs were run.
Conditional forcing, approximate geometry, regional donor assumptions,
unresolved source cases and unproven optimizer convergence retain their stated
qualifications.

The release verifies 41 focused tests on both Ubuntu 24.04 and Windows 2022
with Python 3.12.7, an independently installed wheel, public parameter generation,
exact historical and restart reproduction, independent water balance and sealed
resume. The prior full regression (1,862 passed, one skipped) is retained as
evidence for the unchanged scientific code; it is not described as a new full
regression run. release_evidence.json records the exact source commit,
environments, checks and comparison against 0.3.0.

Use the application source ZIP for orchestration and examples; the wheel supports
frozen-profile simulation. Install dependencies separately. The rendered
documentation and SHA256 checksums are supplied alongside them.

Documentation: https://fuadyassin.github.io/DZTR_model/

License: MIT for software and documentation; third-party input terms remain
applicable. The model paper is a separate citation:
https://doi.org/10.5194/hess-23-3735-2019.

Software DOI (v0.3.1): https://doi.org/10.5281/zenodo.23631633.
The all-versions DOI remains https://doi.org/10.5281/zenodo.23606592.
Zenodo archives the full tagged public source; the wheel, allowlisted application
bundle and rendered documentation are GitHub release assets. Version 0.3.0 and
its DOI https://doi.org/10.5281/zenodo.23606593 retain their original contents.

DZTR 0.3.0

DZTR 0.3.0 provides a researcher workflow for estimating reservoir operating
targets, calibrating them where observations exist, and applying frozen profiles
in hydrological models.

• Data-availability routes for storage and release, release only, storage only,
  and neither response, with documented assumptions and provenance.
• Level-to-storage preparation with explicit datum/reference and geometry choices.
• Wet, dry and normal calibration-year selection with at least 65% support;
  joint NSGA-II, scalar DDS and optional GA calibration; persistent checkpoints.
• Public dztr commands, source orchestration and independently installed-wheel
  application. Frozen examples cover Gardiner, Oahe and Garrison.

Installation, public parameter generation, exact historical/restart reproduction,
water balance and sealed resume are checked on Ubuntu 24.04 and Windows 2022
with Python 3.12.7. The complete local Windows regression suite passed 1,854 tests
with one skip. release_evidence.json identifies the source commit, CI run,
environments and checks. No new calibration searches were needed for this release.

Use the source ZIP for the complete workflow and examples; the wheel supports
frozen-profile simulation. Install Python dependencies separately. The
documentation ZIP is a static-site snapshot; SHA256SUMS.txt seals each file.

Documentation: https://fuadyassin.github.io/DZTR_model/

The four-route assessment compares the same 27 successful, scored reservoirs
across all routes. It documents weaker storage behavior when storage is hidden.
The experiment uses conditional forcing and limited calibration budgets. Software
reproducibility does not establish independent forecast skill, source certification
or optimizer convergence. Regional donors, approximate geometry and source-gated
reservoirs retain their stated qualifications. Private archives are excluded;
only the approved public reservoir examples are distributed.

License: MIT for software and documentation; third-party input terms remain
applicable. Cite this software with CITATION.cff and cite the model paper
separately: https://doi.org/10.5194/hess-23-3735-2019.

Software DOI (v0.3.0): https://doi.org/10.5281/zenodo.23606593

All-versions DOI: https://doi.org/10.5281/zenodo.23606592

Zenodo archives the full tagged public source. All 676 archived files were
verified against the frozen v0.3.0 Git tag. The wheel, allowlisted application
bundle, rendered documentation and checksums are the GitHub release assets.
The original tag and eight sealed files retain their original bytes and hashes.
The software DOI is separate from the model-paper DOI.