Molmer-Sorensen frequency model: improve heuristics for scans not centred on resonance

[AUTProgram]

When we probe two-qubit gate dynamics, we typically start by scanning the gate detuning over a wide range, such that the scan contains the line of symmetry (where mode is driven on resonance). In this case, the MolmerSorensenFreq.estimate_parameters() method works fine because it looks for the symmetry point. However, sometimes we also just scan the gate detuning over a narrower range around the loop closure. In this case, the fit gives garbage, see below for an example.

For now, I'm handling this issue by specifying a user estimate for w_0 manually depending on the scan range. If we can think of a simple heuristic that helps with this, probably worth adding to the estimation procedure directly rather than relying on the user.

[hartytp]

Chatted about this IRL. I think this is a "won't fix" issue. Motivations:

• doing a narrow scan around a specific revival is something one generally only does when one knows the gate detuning relatively accurately so it's easy to supply an accurate user_estimate. As such there isn't a huge amount of value to be had in getting an accurate heuristic here
• I suspect it would be pretty hard to write a general-purpose heuristic which could figure out that we're doing a narrow scan around a specific revival. To make that work, we'd probably have to hard code a lot of assumptions, so this would be a niche special-casing

Given the above, this feels out of scope for ionics_fits and is best handled by users giving the fit an initial guess