Reformat directory format for data downloads

[Fixed in #766 ] 1. Fix data population bug

To reproduce: go to an experiment with multiple survey fields, download data, and inspect. [experiment_name].json has the correct fields, but [experiment_name].participantData.csv is empty, and only contains an empty column for one survey question. Make sure that the correct data is populated in participantData.csv.

2. Reformat directory for ease-of-use

Current directory and file format:

[experiment_name]_Alerts.csv

[experiment_name]_ChatHistory_[CohortID]_[ChatDiscussionName].csv // For each cohort.

[experiment_name]_Logs.json

[experiment_name]_participantData.csv

[experiment_name].json

Proposed directory and file rename / reshuffling:

alerts.csv // Only if there were alerts (don't generate an empty CSV). This is how it works for payouts currently; if there are no payouts enabled, no file is generated.

For each unique chat stage:
[ChatDiscussionName]/ 
    [CohortID].csv
    ...

llmLogs.json 

participantData.csv 

experimentConfig.json 

The data-population portion is marked fixed in #766. Start by locating the data-download generation entry point and comparing its output with the existing payout-file behavior; done means downloads use alerts.csv only when alerts exist, group chat CSVs under chat-stage directories, and use the requested llmLogs.json, participantData.csv, and experimentConfig.json names.