Update README for one-shot runs with pruners that require gradients

[eldarkurtic]

Is your feature request related to a problem? Please describe. It would be a good idea to mention here https://github.com/neuralmagic/sparseml/blob/main/README.md#:~:text=Instead%20of%20training,manager.apply(model) that one-shot pruning runs with pruners that use gradients require passing additional argument (grad_sampler) when calling into manager.apply(model). If grad_sampler not provided, gradients won't be collected and the run will crash.