feature: task-wise agents

as a part of the amped up agentic integration into the app, Grace and Donglai reccomend that we have the following agent split:

1. A project management agent
2. One agent per task (AKA module loaded)

When queried respectively,

1. the project management agent should provide high level insight on what the user should do next in their workflow to advance their project goals
2. the taskwise agents should provide more granular insight on how to go about doing a given task in a module (i.e. training a model might seem quite intuitive to us, but is certainly not for a user)

substantial ideation on this thread should be done, but some key implementation steps should likely be:

1. Spawning of an initial pool of agents on app startup (not just the one)
2. Creation/configuration of context to provide to respective agents

The crux of this task should be the creation of the taskwise agents, not the refinement of the existing project management agent

No files, tests, or entry points are named. Start by tracing app startup and the existing project management agent, then determine how task/module context should reach separate agents; done means task-wise agents can provide granular guidance while the project management agent retains its high-level role.