Fix panic in DecodeStream::step due to incorrect index usage

[n0gu-furiosa]

When calling DecodeStream::step multiple times, it eventually panics with attempt to subtract with overflow in the following lines of code:

tokenizers/tokenizers/src/tokenizer/mod.rs

Lines 1108 to 1109 in 24d29f4

	let new_prefix_index = ids.len() - *prefix_index;
	*ids = ids.drain(*read_index..).collect();

The panic can be easily reproduced, and I have added a test case to demonstrate the issue.

Upon inspecting the code, I found that the shrinking of the token buffer references read_index instead of prefix_index. This PR corrects the issue by using the correct index.

However, this change makes read_index unused, so I am not entirely certain if it aligns with the intended logic of the original implementation. Please let me know if further adjustments or clarifications are needed, or if there is additional context regarding the intended use of read_index.