CSV Label Formatting

Hello, I was looking through the metadata and wanted to clarify the labeling used in `emg2pose_metadata.csv`.

1. The first column is named `session`, however in session `2022-04-07-1649318400-8125c-cv-emg-pose-train@2`, there is data that is not contiguous in time (there are other sessions that occur between stages of this session). The first continuous part of this session starts at `1649314883.5666700` and the second continuous part of this session starts at `1649398329.3833300`.

2. It also seems like it is never defined what `demonstration` and `training` mean in the session label.

3. This is just a small detail, but for some users, there seem like there are 2 sessions or 5 sessions, although the paper mentions users only have 3-4 sessions each. These 2 sessions seem to be the only sessions from another user: `2022-12-06-1670313600-e3096-cv-emg-pose-train@2`, `2022-12-06-1670313600-c3e72-cv-emg-pose-train@2`. These 5 sessions seem to be from one user: `2022-04-13-1649836800-a5d25-cv-emg-pose-train@2`, `2022-04-13-1649836800-45bfc-cv-emg-pose-train@2`, `2022-04-13-1649836800-8892f-cv-emg-pose-train@2`, `2022-04-13-1649836800-e6cfb-cv-emg-pose-demonstration@2`, `2022-04-13-1649836800-6e2c7-cv-emg-pose-demonstration@2`.

Start by inspecting emg2pose_metadata.csv and the session-label definitions in the paper or repository documentation. Verify the reported discontinuities, session counts, and the meanings of “demonstration” and “training”; done means the metadata labeling is clarified or corrected and the observed discrepancies are explained.

Written by the indexing model from the issue text.