FIX Mantis solver: align with updated Linear Probe Adapter

[bryan29-ly]

PR #34 updated the benchmark_utils/adapters/linear_probe.py but solvers/mantis.py was not completely adapted:

• run() still passed max_iter=self.max_iter → crash.
• run() read self.n_estimators, but the parameter was declared as
  n_iterators
• penalty, C and alpha were declared in parameters but never forwarded
  to the adapter.

Fix

• Remove the max_iter argument.
• Rename the n_iterators parameter to n_estimators
• Forward penalty, C and alpha to the Linear Probe Adapter

cc @ambroiseodt

[ambroiseodt]

Seeing this late @bryan29-ly, seems great! Thanks @felixdivo for merging it :)