MicrosaccadeHelper not re-initializing self._number_of_trials

**Problem**: Currently the `MicrosaccadeHelper` does not re-initialize `self.number_of_trials` after a model `candidate` calls `start_task(require_variance=True, number_of_trials=>2)`. This is a problem when in fitting microsaccades are wanted, but during `candidate.look_at()` microsaccades are not wanted (i.e., `candidate.look_at(require_variance=False, number_of_trials=1)`). This results in an error in `core.py` `_package_layer`, as the expected dimensions do not match (`number_of_trials` is read out incorrectly).

**Workaround**: when calling microsaccades on the fitting stimuli, but not the testing stimuli, call `candidate.look_at(require_variance=True, number_of_trials=1)` to correctly set the expected `number_of_trials`.

**Solution**: always re-initialize `MicrosaccadeHelper.number_of_trials` regardless of whether `require_variance` is `True` or `False`, or other similar approach.

Start by locating MicrosaccadeHelper and tracing start_task(require_variance=...) alongside candidate.look_at() calls. Check how core.py's _package_layer reads number_of_trials; done means the helper re-initializes that value for both variance settings and the expected dimensions no longer mismatch in the described fitting/testing sequence.

Written by the indexing model from the issue text.