[Copilot]

The Python API could only load data from files, which forced users who already hold a DataFrame to write it to disk first. This PR lets Application take a DataFrame directly and optionally rename its columns via a mapping, so users don't have to change their own frames.

• Application
  • load_dataframe(df, column_mapping=None) loads a DataFrame as the working data.
  • get_dataframe(column_mapping=None) returns the current data as a DataFrame. It raises TypeError if the loaded data isn't DataFrame-backed (e.g. spectra loaded from MSP).
• DataFrame data class
  • load_dataframe and get_dataframe hold the logic. Both work on a copy, so the caller's frame is never mutated.
  • load_dataframe raises TypeError for non-DataFrame input.
  • column_mapping is {source_name: target_name}. Columns not in the mapping are kept as they are.
• Tests: tests/test_DataFrame_api.py covers load with mapping, no mutation of the input, and the reverse rename.

app = Application()
app.load_dataframe(df, {"Name": "compound_name"})
await app.annotate_spectra(converters=["PubChem"])
out = app.get_dataframe({"compound_name": "Name"})  # back to the user's own column names

Not covered:

• There is no validation that the required columns are present.
• The README, docs and CHANGELOG are not updated for the new methods.

• Fixes Improve Python API to accept DataFrames for direct use #174