Add planning mode for AI Assistant

[philippjfr]

Up until now the AI Assistant was fairly dumb, it simply tried to pick the agent that would actually satisfy the user query and then resolved which "dependencies" would have to be satisfied to give that agent all the context it needed. Apart from potentially requiring a whole number of calls to the LLM to satisfy all the dependencies this had the drawback that it doesn't actually plan the query, so if some intermediate steps are needed it would never perform them. In the new planning mode it takes a different approach, it gives the LLM the chance to break down the user query into a series of steps each of which will be performed by an agent, the assistant makes a plan to solve the task, gives each agent an instruction on what task it should perform and then we validate that the plan it made correctly provides all the context that is needed. If the plan is invalid we ask it to re-solve it, now with the information on where it went wrong.

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 133 lines in your changes missing coverage. Please review.

Project coverage is 55.40%. Comparing base (2de7246) to head (ea73d71).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
lumen/ai/assistant.py	0.00%	112 Missing ⚠️
lumen/ai/models.py	0.00%	15 Missing ⚠️
lumen/ai/agents.py	0.00%	4 Missing ⚠️
lumen/ai/__init__.py	0.00%	1 Missing ⚠️
lumen/ai/utils.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #719      +/-   ##
==========================================
- Coverage   55.77%   55.40%   -0.38%     
==========================================
  Files          98       98              
  Lines       11896    11976      +80     
==========================================
  Hits         6635     6635              
- Misses       5261     5341      +80     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.