[0xadvait]

Adds abliteration.ai as a provider (OpenAI-compatible, https://api.abliteration.ai/v1) and registers Abliterated Large v2 (abliterated-model-large-v2, 1M context, text in).

• New abliteration_api_key, injected via /v1/keys like the others. run-enclave.sh reads ABLITERATION_API_KEY from the env file. The key is not in the image.
• Pricing: $3.00 input, $5.00 output per 1M tokens, from https://abliteration.ai/pricing and the provider's /v1/models (prompt 0.000003, completion 0.000005).
• Params: /v1/models lists temperature (0 to 2) as supported, so no restriction flags. It's a reasoning model; reasoning tokens arrive in usage.completion_tokens_details.reasoning_tokens and are billed as output.
• Tests: resolve, aliases and cost in test_pricing.py; added to the live provider suite (needs an ABLITERATION_API_KEY repo secret).
• Verified: make lint and the CI test command pass (503 passed). A live non-streaming and streaming request through chat_controller returned content, usage and the expected cost.

Ops: add ABLITERATION_API_KEY to the enclave env file and repo secrets before deploying. No dependency or uv.lock change.