Consider normalization of log levels passed to set_log_level()

The #597 implements configuring specific libssh log levels, but it would be good to be able to accept arbitrary values from the python logging system range and normalize to closest value somehow.

Looks like this is untested. But also, it might make more sense to allow arbitrary levels to be normalized to the closest known value as the Python logging infra does not have a problem with intermediate logging levels as long as they are integers.

_Originally posted by @webknjaz in https://github.com/ansible/pylibssh/pull/597#discussion_r2813078766_

Start at the set_log_level() implementation and review the configuration work from PR #597, then compare its accepted values with the integer range used by Python logging. Check the existing test coverage, which the issue says is missing. Done means arbitrary integer levels are consistently normalized to the closest supported libssh level and covered by tests.

Written by the indexing model from the issue text.