feat(langchain): emit streaming latency span attrs

[sayonfortify]

Adds FortifyRoot streaming latency span attributes to LangChain LLM spans when LangChain emits streaming token callbacks.

• Records fortifyroot.llm.streaming.time_to_first_token_ms on the first on_llm_new_token callback.
• Records fortifyroot.llm.streaming.time_to_generate_ms at on_llm_end, only when a first token was observed.
• Leaves non-streaming and empty/no-token calls without these attrs, preserving NULL/no-data behavior downstream.
• Adds deterministic callback-level tests; no provider calls or VCR cassette.

Local validation:

• python3 -m py_compile ...
• PYTHONPATH=packages/opentelemetry-instrumentation-fortifyroot /Users/sayondutta/opt/anaconda3/envs/py3.12/bin/python -m pytest packages/opentelemetry-instrumentation-langchain/tests/test_streaming_latency_span_attrs.py -q -> 2 passed
• git diff --check clean
• secret scan clean