fix: correct stop_sequence -> stop_sequences typo in _continuous_greedy_until

[RudrenduPaul]

Closes #888

What this PR does

Fixes a typo in TransformersModel._continuous_greedy_until where the code
read split[0].stop_sequence (singular) instead of split[0].stop_sequences
(plural). The Doc/request object only defines a stop_sequences field
(src/lighteval/tasks/requests.py), so the singular access either raises an
AttributeError or silently returns None depending on the object type,
breaking stop-sequence handling for continuous batching generation.

Root cause

src/lighteval/models/transformers/transformers_model.py, inside
_continuous_greedy_until, builds stop_tokens from the first item in the
batch:

# before
stop_tokens = split[0].stop_sequence

# after
stop_tokens = split[0].stop_sequences

The correct attribute name (stop_sequences) is already used consistently
elsewhere in the same file (e.g. _generate uses
batch[0].stop_sequences), confirming this was a one-off typo rather than
an intentional different field.

Change

Single-line fix in src/lighteval/models/transformers/transformers_model.py
(line 561). No other file references the typo'd singular form, so this is
the only place that needed updating.

Verification

• Confirmed stop_sequences (plural) is the actual field defined on the
  request/Doc object in src/lighteval/tasks/requests.py.
• Confirmed via grep that no other code path in
  transformers_model.py references the singular .stop_sequence typo.
• ruff format --check and ruff check both pass on the changed file.

Note: Claude Code was used to assist in drafting this fix. All changes were reviewed by the submitter.