[ENH] - Upate / tweaks to the outlier flattening in AP fitting

[TomDonoghue]

During reviews for the Developmental Demo paper, we got some feedback about how the current approach for flattening outliers in the aperiodic fitting might be sub-optimal for some development data.

In their suggested edit, the outlier flattening in _robust_ap_fit was updated to:

if min(flatspec) < 0:
    flatspec -= min(flatspec)

Correspondingly, the fit function was edited to add the following line to line 458:
self._spectrum_flat[self._spectrum_flat < 0] = 0

This is something to explore and we should build in this customization to the next release.