fix(edgelist): rename legacy marker columns in record batches

[Aratz]

Summary

• Edgelist.to_record_batches renames legacy marker1 and marker2 columns to marker_1 and marker_2, matching to_polars, to_df, and iterator.
• Independent of the panel-concatenation stack. Base branch is dev.

Test plan

• CI passes (ruff, ruff format, mypy, pytest)
• test_record_batches_rename_legacy_marker_columns covers a legacy record batch

Made with Cursor

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 7cf60ff. Configure here.}

[Aratz]

@elhb I've done a couple changes here:

1. instead converting each batch to a dataframe to rename the columns and then convert it back to a record batch, I adjusted the SQL query so that this is done immediately without any extra conversions or copies
2. Eventually, such backward compatibility patches will accumulate, which will both affect readability and make it difficult to get an overview of such patches if we decide to deprecate them. I've started to isolate such patches in a legacy/ folder. This way the patches don't stand in the way as much and it becomes much easier for us to audit and possibly remove such patches later.