chore(weights): replace sparkinfer-k3 with sparkinfer

[e35ventura]

Summary

• Remove gittensor-ai-lab/sparkinfer-k3 from the repository registry.
• Add gittensor-ai-lab/sparkinfer with emission_share set to 0.3.
• Carry over the former K3 eligibility, label, base-score, issue-discovery, maintainer-cut, and trusted-label settings.
• Remove the K3-specific scoring.time_decay override so SparkInfer uses the platform default time decay.

Emission share

Repository	Before	After
gittensor-ai-lab/sparkinfer-k3	1.0	Removed
gittensor-ai-lab/sparkinfer	Not registered	0.3

Why

This moves Gittensor incentives from the K3-specific repository to the main SparkInfer repository and applies the requested 0.3 emission allocation. The configuration incorporates the repository owner's PR feedback while returning time-decay behavior to the platform defaults.

Validation

• jq empty gittensor/validator/weights/master_repositories.json
• uv run --extra dev pytest tests/validator/test_load_weights.py (64 passed)

[ai-hpc]

On hyperparameters: rather than dropping to platform defaults, could the sparkinfer entry keep the same hyperparameters sparkinfer-k3 had (eligibility, fixed_base_score, issue_discovery_share, label_multipliers, maintainer_cut, trusted_label_pipeline, default_label_multiplier), just under the new repo name with emission_share: 0.1?

The one exception is scoring.time_decay — drop that block entirely so the repo uses the platform default time decay instead of sparkinfer-k3's custom curve.