Update Pandora track and shower discrimination scheme and training

[rtriozzi]

This PR changes the Pandora XML (particularly, the LArBdtPfoCharacterisation algorithm) to accomodate for changes to the track/shower discrimination scheme, and to introduce the corresponding new BDT training files.

---

Context

The code change within Pandora consists in choosing the Collection plane to compute calorimetric variables used in the track/shower discrimination BDT, instead of relying on the Induction-1 plane. Due to mapping technicalities, the Pandora W view (commonly mapped to Collection in other LArTPCs) is mapped to Induction-1 in ICARUS, whereas the U and V views are mapped to Induction-2 or Collection, based on the TPC. For more details on the Pandora code change, check the corresponding PandoraPFA/LArContent PR #246, SBN DocDB 40882, and SBN DocDB 40375.
The UseICARUSCollectionPlane algorithm parameter controls the plane used to determine whether charge information is available or not for the reconstructed particle (Collection if true), and as a result which BDT to use.
Two ICARUS-specific tools are defined: ConeChargeFeatureTool_ICARUS, and ThreeDChargeFeatureTool_ICARUS. The UseICARUSCollectionPlane tool parameter controls the plane used for computing charge-based variables (Collection if true).
This modification was agreed with Pandora experts.

---

BDT re-training

Information on the BDT re-training are available on SBN DocDB 42432.
The track/shower BDT was re-trained with BNB MC samples based on icaruscode v09_84_00_01, balanced between track and shower particles, and with "good reconstruction" cuts. The accuracy of the classification is shown to improve substantially across all particle types.

---

Dependency

This PR needs the LArSoft/larpandoracontent v04_16_00 release (LArSoft/larpandoracontent PR #78), i.e., larsoft v10_08_02.
It also needs the corresponding BDT XMLs added to icarus_data: a new icarus_data release will be needed to pick up the right XMLs.
After this PR is merged, the CAFMaker needs to be updated to pick up the newly-named tools. A companion PR in sbncode was opened (sbncode PR #546).

---

Review

Tagging for review:

• @brucehoward-physics and @cerati as reconstruction experts and shower reconstruction task force conveners;
• @SFBayLaser for assistance with icarus_data and LArSoft releases.

Also tagging @acampani as assignee.

Thanks!

[brucehoward-physics]

Just curious - why are these in a separate file instead of putting both into one file as before? (wondering if just a choice or if there is a particular reason e.g.)

Also: checking, in the doc-db presentation you've given with the training results, were they prepared with these changes, so in some way this is already validated?

[rtriozzi]

The two BDTs (with charge information and without it) came from two separate trainings, so I saved them into two separate XMLs. If you think it would be nicer to have them within a single XML, I can merge them! Thanks Bruce!

[brucehoward-physics]

I think it's probably fine either way.