Improve Data Generation stage

[nkarpovdb]

Our data generation stage is currently a python job configured with a JSON config file.

Some improvements to consider:

1. Move this to a Databricks app that has its own custom UX to control data generation
2. Decouple static data (menus/items) so that each location can have its own config. Right now all locations share the same menus.

[nkarpovdb]

Some more thoughts on how we should handle data generation.

Today we have this long running job that couples both config and generation. This is neither realistic nor does it allow us to easily extend the generated events.

A strawman proposal:

1. We create a Databricks App for ordering, this should literally be like UberEats, DoorDash etc. We can likely vibecode this very fast.
2. This app generates the first order_created JSON and writes it to the events volume
3. A separate process (Spark Streaming, real-time!?) watches for order_created events and generates the events for the remainder of the order lifecycle.

This kind of structure still supports data generation as we have it today (just a script to do create orders using the Apps API?), but also allows us to hook into the flow: anyone can write that first order_created event. So, in the future, an MCP server could control volume, etc.

We'd still need a way to better configure locations, menus, items, etc. as right now they are coupled too. Maybe this app also handles configs.