remove force-multi and fix None val check in base tuner in autotuning

[cli99]

This PR fixes the following bugs in autotuning:

1. the autotuning scheduler does not add force-multi to the experiment script and uses what users pass in
2. the initial best metric value could be None in the tuning and needs to be checked

[rocm-mici]

Can one of the admins verify this patch?

[loadams]

These changes were already applied in other PRs, closing this one.