[State Service] Scaffold position estimation and decide on approach

[TKDPenguin]

Goal:
The week 1 spike researched methods for estimating car position from noisy, multi-sensor data. This ticket closes that spike by making a documented decision on the approach, then scaffolding the implementation so week 3 can go straight into building rather than designing. The scaffold itself does not need to produce good position estimates. It just needs to define the structure that the real algorithm will use.

Technical Description:

• Review the spike findings as a team and decide on a primary approach and document the decision along with the reasoning in the repo (a short markdown file in the repo is fine)
• Scaffold the state estimation service: define the class/function structure, inputs (cleaned wide DataFrame from the clean service), and outputs (positions DataFrame matching the positions data model)
• Define clearly what the service needs from the clean service (which signals, at what sample rate). This becomes an implicit contract between the two
• Identify any external libraries needed and add them to the project dependencies

Conditions of Satisfaction:

• The approach decision is documented in the repo with a brief rationale
• The state estimation service accepts a cleaned DataFrame and returns a positions DataFrame matching the defined data model
• Required external libraries are added to dependencies and confirmed to install correctly