[DimRed] True features baseline evaluated twice

[lazappi]

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the main branch of openproblems.

---

I think that now that the data loaders return logCPM normalised expression in adata.X the true_features and true_features_logCPM baseline methods for the dimensionality reduction task are now identical and we should only use one of them (or modify true_features to use raw counts?).

[scottgigante-immunai]

You're right -- all of the true_features except the logCPM version can/should be removed -- but I'd like to check that it's performing as expected in the next run before I do so.

[scottgigante-immunai]

Closed in #816