S2025 perception cone detection final PR

[Ren990420]

Perception teams final PR

1. Cleaned up the code. Move more variables to yaml file to let user have the chance to change them.
2. Add a README.md in knowledge/detection about where to download different detection models if people do not want to use the default one.
3. Add the pedestrain_detection.py and its corresponding yaml file. This file was not used by any verticals; however, it was used by planning horizon group for yielding test. Directly adapted from cone_detection but change the output to AgentState and add velocity recording.

[sonarqubecloud]

Quality Gate failed

Failed conditions
5.3% Duplication on New Code (required ≤ 3%)

See analysis details on SonarQube Cloud

[krishauser]

LGTM