chore(weights): configure Gittensor-TinyRouter maintainer cut, eligib…

[James-CUDA]

Summary

Activates a full reward configuration for James-CUDA/Gittensor-TinyRouter. Previously the entry carried only emission_share / issue_discovery_share and inherited all global defaults. This sets an explicit per-repo policy that opens up eligibility, shortens the reward window/decay, introduces label multipliers (including a king reward), and carves out a 30% maintainer cut.

Changes

The config mirrors the established mini-router/minirouter policy.

Field	Before (global default)	After
maintainer_cut	0.0	0.3
eligibility.min_valid_merged_prs	3	0
eligibility.min_credibility	0.80	0.0
scoring.pr_lookback_days	30	7
scoring.time_decay.sigmoid_midpoint_days	10	3
label_multipliers	(none)	bug 0.1, feature 0.15, refactor 0.02, king 10.0

Rationale

• Eligibility relaxed (min_valid_merged_prs 3→0, min_credibility 0.80→0.0): new contributors can earn rewards immediately rather than clearing the global 3-PR / 0.80 credibility gate.
• Shorter reward window (pr_lookback_days 30→7) and faster decay (sigmoid_midpoint_days 10→3): rewards concentrate on recent, relevant work and stale PRs depreciate quickly.
• Label multipliers: scale PR value by type — bug/feature/refactor are de-emphasized, while a king label grants a 10× reward for high-impact contributions.
• Maintainer cut 30%: routes a share of the repo's emission to registered maintainer miners via the existing maintainer_cut carve-out.

Validation

• All values within the loader's enforced bounds (maintainer_cut ∈ [0,1], min_credibility ∈ [0,1], min_valid_merged_prs ≥ 0, pr_lookback_days ∈ [1,90], sigmoid_midpoint_days ∈ [1,90], label_multipliers ≤ 10 entries / values ∈ [0,20]).
• Total emission_share across the registry remains 1.0 (this PR does not alter emission_share).
• load_master_repo_weights() parsing and all three validators (_validate_emission_shares, _validate_eligibility_configs, _validate_scoring_configs) pass for this entry.

Verification

• py -m pytest tests/validator/test_load_weights.py

[James-CUDA]

hi, @anderdc, plz can you review PR.

[James-CUDA]

thank you