Minor improvements

[jdrugo]

I suggest a range of minor improvements to the current version of the InfluenceCalculator class:

• Add count_thresh=5 argument to __init__ that in turn is provided to _load_sql_data to support customized thresholds on the synapse counts.
• Clarify in docstring possible values to syn_weight_measure argument of _create_sparse_W method.
• Clarify in docstring that the silenced_neurons argument of the calculate_influence method needs to be a list or numpy array of neuron_ids rather than a pandas table, and update the usage example accordingly.
• Eventually turn the _solve_lin_system method into a static method, as it doesn't seem to modify any class variables.
• Ensure that the _normalize_W method directly modifies W_norm, such that it doesn't need to be returned, and that W_norm = self._normalize_W(W_norm) can be replaced by a simple call to self._normalize_W(W_norm).
• To clarify that the _set_columns_to_zero expects silenced_neurons to be W matrix indices rather than neuron_ids, rename argument to silenced_W_idcs and update the method's docstring to reflect that.

[ZakiAjabi]

• Add condition for connectivity matrix rescaling such that it only applies if the largest eigenvalue of the original matrix is larger than 1.