[Docs] Consolidate environment variables reference in docs/configuration.md

[ChaitanyaAIML-lab]

Summary

Consolidate 31 environment variables across the project into a single reference table in docs/configuration.md, with cross-references from README.md, docs, and evals.

Closes #1

Why

Environment variables were previously documented across multiple locations (.env.example, README.md, docs/decision-models.md, docs/skills.md, and evals/README.md). New contributors and users had no central reference to view all variables, default values, and reader subsystems.

What Changed

• Created docs/configuration.md with a structured 4-column reference table (Variable, Who reads it, Default, What it does) documenting all 31 environment variables.
• Added docs/configuration.md to the primary Docs listing in README.md.
• Replaced per-page variable documentation in docs/decision-models.md, docs/skills.md, and evals/README.md with links to the configuration reference.
• Added entry to CHANGELOG.md under Added.

How to Verify

1. Run code formatting check:
   ruff format --check .

2. Run lint check:
   ruff check .

3. Run type check:
   ty check

4. Run the unit test suite:
   pytest -q

[Yunaik]

Thanks for picking this up and following the issue closely. The table has the right shape and most rows match the code. A few fixes before merge:

1. The table lists TYPESAFE_API_URL as "(unset)". In code it defaults to https://openrouter.ai/api/alpha/decisions (s1a/decision_models/wire.py:20). The direct TypeSafe endpoint in the TYPESAFE_API_KEY row is https://api.typesafe.ai/v1/systemone (wire.py:22).
2. LAYA_DEVICE, LAYA_MAX_LEN and LAYA_HEAD_MAX_LEN have no default in code. laya.py:101 passes None for the device and the Laya library picks one; the two lengths override the checkpoint's window only when set (laya.py:114). The values in .env.example are examples. Please write "(library default)" and "(checkpoint default)" for these.
3. The issue also asks to cut the per-page repeats down to a pointer. Each variable is now documented in two places: the pages keep their variable lists and also gain a link. Keep what a reader needs to act at that spot (for example, which key to export in docs/skills.md) and move the rest into the table.
4. [Refactor] Rename the --slot flag to --model #7 renames --slot to --model. Please rebase once it lands and update the "Who reads it" column to match (for example, jev model). docs/decision-models.md will conflict.
5. PLAYWRIGHT_MCP_COMMAND appears only in the usage docstring of evals/replay/cast.py. Please check which process reads it and name that in the column, or drop the row.

Happy to merge once these are in.

[ChaitanyaAIML-lab]

Thank you for the detailed review and guidance, @Yunaik!

All requested changes are now in place and rebased on latest main:

1. TYPESAFE_API_URL & TYPESAFE_API_KEY: Set the default for TYPESAFE_API_URL to https://openrouter.ai/api/alpha/decisions (wire.py:20) and documented the direct TypeSafe endpoint as https://api.typesafe.ai/v1/systemone (wire.py:22).
2. LAYA_* Defaults: Updated LAYA_DEVICE to (library default) and LAYA_MAX_LEN / LAYA_HEAD_MAX_LEN to (checkpoint default).
3. Trimmed Per-Page Repeats: Reduced per-page repetitions in docs/skills.md, docs/decision-models.md, and evals/README.md down to pointers to docs/configuration.md.
4. Model Flag Alignment: Updated the "Who reads it" column to reference models (e.g. jev model, laya model) instead of slots.
5. PLAYWRIGHT_MCP_COMMAND: Attributed to scripts/browser_showcase.sh with default node.

Verified locally with ruff format, ruff check, ty check, and the test suite (pytest -q, 482 passed). Ready for your review!

[Yunaik]

thank you