Revert "[lora][tinker] Add pause and resume for multi-tenant lora "

[erictang000]

Reverts #1657, since we can skip pause/resume for multi-tenant lora after #1677

[gemini-code-assist]

Code Review

This pull request removes the transient per-LoRA targeted pause and resume functionality, including the lora_name parameter from generation control methods and the sample_with_retry logic. Associated server endpoints and test suites have also been deleted. Feedback identifies a potential issue in the sample method where removing dictionary guards for prompt and sampling_params could lead to AttributeError if those fields are explicitly provided as null.

[gemini-code-assist]

The removal of the or {} guard for prompt and sampling_params reduces the robustness of the sample method. If the input JSON explicitly contains null for these fields, body.get(key, {}) will return None, leading to an AttributeError when .get() or other dictionary methods are called on them later (e.g., at line 635).

Suggested change

	prompt = body.get("prompt", {})
	# Render prompt: flatten text tokens and, if images are present,
	# call the render endpoint to get placeholder tokens + features.
	token_ids, mm_features = await self._render_for_sample(prompt, session_id, model=model)
	return await self._sample_with_rendered_tokens(
	token_ids=token_ids,
	mm_features=mm_features,
	body=body,
	session_id=session_id,
	model=model,
	)
	async def _sample_with_rendered_tokens(
	self,
	token_ids: List[int],
	mm_features: Optional[MultiModalFeatures],
	body: Dict[str, Any],
	session_id: Optional[str],
	model: str,
	) -> SampleResponse:
	"""Dispatch a single rendered sample request to /inference/v1/generate.
	Shared between ``sample`` (one shot) and ``sample_with_retry`` (loops
	on this method with accumulating ``token_ids`` and decremented
	``max_tokens``). No retry, no LoRA gating; those live in the caller.
	"""
	num_samples = body.get("num_samples", 1)
	tinker_params = body.get("sampling_params", {}) or {}
	tinker_params = body.get("sampling_params", {})
	prompt = body.get("prompt") or {}
	num_samples = body.get("num_samples", 1)
	tinker_params = body.get("sampling_params") or {}