feat: Support for setting aggregation for (Hybrid)RouteLayer

[andreped]

fixes #201.

Also fixed:

• Added unit tests to verify that aggregation selection updates the aggregation
• Added unit tests to verify that aggregation works as intended for both RL and HybridRL
• Added support to set top_k for RouteLayer similar to HybridRouteLayer
• Added unit tests for top_k for RouteLayer
• Added some error handling when setting both top_k and aggregation
• Linted code

[codecov]

Codecov Report

Attention: Patch coverage is 83.33333% with 5 lines in your changes are missing coverage. Please review.

Project coverage is 78.00%. Comparing base (1394adf) to head (7cd3755).

Files	Patch %	Lines
semantic_router/layer.py	82.35%	3 Missing ⚠️
semantic_router/hybrid_layer.py	84.61%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #202      +/-   ##
==========================================
+ Coverage   77.96%   78.00%   +0.04%     
==========================================
  Files          42       42              
  Lines        2119     2146      +27     
==========================================
+ Hits         1652     1674      +22     
- Misses        467      472       +5     

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

[andreped]

@jamescalam Just realized that the codecov.xml file is inside the repo. IMO, it does not make any sense to have there. Code coverage is viewed in an external website (see here), which then is linked to this badge in the README, right?

So can it be deleted? I can do this now in this PR, if of interest, as well as add codecov.xml to .gitignore.

[andreped]

I also see that codecov complains about 5 lines above, but they are all related to error handling. Do I need to add these in tests or should I add #nopragma or similar?

[bruvduroiu]

Thank you for the contribution! I agree that only using sum tends to favor Routes with more utterances.

Very small nitpick on the choice of naming the aggregation strings. We want to keep in line with the Pandas lowercase aggregation function names.

[andreped]

Just ran tests with this in one of my projects. Works fine! :] But of course, as a top_k of 2 or 3 worked best for my project, the aggregation method was less important. Then again for higher values of top_k I got a change in performance.