Make replanning efficient

[Yadunund]

Feature request

Description

Each time the optimal_solve/greedy_solve methods are called, the Request::Model object for each request is initialized during which the invariant battery drain and invariant duration of the request are computed. This makes replanning slow as requests that were planned for before will still need their Models to be generated each time. To make replanning faster, we should maintain a cache of Request::Model within the TaskPlanner

Implementation considerations

Refer to comment #15 (comment)