Summary

This PR adds an optional market sentiment feature enricher for existing OHLCV datasets.

It does not add a new price data source. Instead, it provides a small helper that merges structured, lagged sentiment features onto an existing FinRL dataframe so those features can be appended to tech_indicator_list and used as part of the RL state.

What is included

• finrl.meta.preprocessor.adanos_sentiment.add_adanos_market_sentiment(...)
• a small example script showing how to enrich a Yahoo Finance dataset before training
• targeted unit tests
• short documentation updates in the README, examples, FAQ, and data layer docs

Design choices

• Optional and fail-open: without ADANOS_API_KEY, the helper is a no-op and returns the original dataframe
• No provider changes: Yahoo/WRDS/Alpaca remain unchanged
• Lagged features only: sentiment columns are shifted before joining to avoid same-day leakage
• Recent-window positioning: intended for short-horizon or recent-window research where structured market sentiment is available

Example features

The helper adds lagged aggregate and source-level features such as:

• adanos_buzz_mean_lag1
• adanos_sentiment_mean_lag1
• adanos_source_coverage_lag1
• adanos_reddit_buzz_lag1
• adanos_x_buzz_lag1
• adanos_news_buzz_lag1
• adanos_polymarket_buzz_lag1

Validation

Ran locally:

• python3 -m pytest unit_tests/preprocessors/test_adanos_sentiment.py -q
• python3 -m py_compile finrl/meta/preprocessor/adanos_sentiment.py examples/FinRL_StockTrading_2026_1_data_with_adanos.py unit_tests/preprocessors/test_adanos_sentiment.py
• git diff --check

I also tried python3 -m pytest unit_tests/preprocessors -q, but that currently fails in this environment because upstream tests import finrl top-level and require optional dependencies such as gymnasium during collection.