Planning Notes

@XX-Yin @hanhou @rachelstephlee starting the planning process here

Goal:
- This repository is for analysis that uses more than one session of data. In general, `data-utils` and `basic-analysis` should be single-session repositories
- This repository should NOT contain code to run the infrastructure on Code Ocean, that should live elsewhere: https://github.com/orgs/AllenNeuralDynamics/projects/113

Proposed structure:
- Operate on tidy dataframes with a column that identifies the session. In my initial PR (https://github.com/AllenNeuralDynamics/aind-dynamic-foraging-multisession-analysis/pull/1) I used `ses_idx` because this is computed in `data-utils`. Open to something more verbose like `session_index`.
- If we have multiple dataframes (like FIP, and df_trials), then timestamps should be consistent between the two dataframes on a session by session basis. Timestamps between sessions are meaningless.
- Do not use an "analysis object" like the NWB object to pass to functions
- We should strive to limit the number of data objects used in an analysis. That is, add columns to dataframes, rather than also passing in lists or dictionaries with metadata.
- Separate functions that aggregate data from sessions and functions that analyze/plot aggregate data.
- When possible use single-session functions from `data-utils` and `basic-analysis`

Open questions:
- anything we should do to facilitate use of the first-order analysis architecture?
- I already referenced a use case of triggering jobs by subject ID: https://github.com/AllenNeuralDynamics/aind-physio-arch/issues/668#issuecomment-2805899841