Refactor evaluation code for the spiking network

[Sallyliubj]

• Remove AbstractSpikingTask (spiking/tasks.py) to avoid redundancy
• Add Spiking Evaluator class for each task in spiking/eval_tasks.py
• lambda_grid_search now reads a single model path instead of all model paths in one folder
• Replace hardcoded values in evaluate_single_trial (eval_end, eval_amp_thresh) with variables