Draft sampling uses a different distribution than speculative rejection

## Problem

`sample_tokens` samples draft tokens through the native-dtype logits path, while speculative rejection/correction uses the float32 distribution produced by `logits_to_probs`.

In speculative decoding, the proposal sample must be drawn from the same draft distribution used in the acceptance ratio and correction distribution. If the sample is drawn from `r` but the algorithm computes acceptance as if it came from `p`, the final marginal distribution is generally biased.

## Impact

With bfloat16 logits, the sampled draft distribution can differ from the float32 draft distribution used for rejection. That breaks the exact-distribution guarantee expected from speculative decoding.

## Fix

Fixed in PR #30 by casting logits to float32 before temperature scaling and softmax in `sample_tokens`, so sampling and rejection use the same draft distribution.

Review PR #30 alongside the sample_tokens and logits_to_probs entry points. Confirm that draft sampling and speculative rejection use the same float32 distribution, and verify the existing speculative-decoding checks for the exact-distribution guarantee.

Written by the indexing model from the issue text.