Allow tuple in param_scheduler

Context:
I'd like to follow Leslie paper and change the momentum in adam at the same time I am changing the lr.
Issues:
The momentum in adam is equivalent of beta 1 which is the first values of the tuples in the torch initializer:
Adam(params, lr=0.001, betas=(0.9, 0.999), eps=1e-08, weight_decay=0, amsgrad=False) (ie. this is 0.9). However, right now there is no way to use param_scheduler with tuples let alone specifying only one tuple.
Proposed solution:
Allow param_name to be a bit more complex where you can specify things such as betas[0] (or something in the line of what pydash is doing pydash.get(param_gp[0],"betas[0]") which would require to basically use base_get) that will extract the first tuple from the betas.
I think it might be able to be done in the __call__ method in ParamScheduler

Start in ParamScheduler.call and inspect how param_name currently resolves optimizer parameters. Define the supported tuple-path behavior described in the issue, including betas[0] and changing it alongside the learning rate, then verify that scheduling one tuple element leaves the other values unchanged.