fix(models): add Gemini 2.5 token limits

[amirshahzadhashmi7145]

Summary

• Fixes the leftover part of Unknown model_tokens silently falls back to 8192 and truncates — warning goes to stderr, once per process, and never reaches the result #1121: google_genai/gemini-2.5-flash (and related aliases) were missing from the token table, so the graph silently used an 8192-token window and truncated long pages.
• Registers Gemini 2.5 / gemini-flash-latest with a 1M context on google_genai and google_vertexai.

Test plan

• uv run pytest tests/test_models_tokens.py::TestModelsTokens::test_gemini_2_5_models_are_registered -q
• uv run black --check scrapegraphai/helpers/models_tokens.py tests/test_models_tokens.py
• Confirm a graph with google_genai/gemini-2.5-flash and no explicit model_tokens gets model_token == 1000000 (not 8192)