Show an error when attempting to re-train a synthesizer that was created on a previous SDV version

[npatki]

Problem Description

Consider the following scenario:

1. I am on the latest version of SDV (1.10.0) and I encounter a bug
2. I am able to find a workaround so I move on from the bug. I train and save a synthesizer on SDV 1.10.0
3. In a future SDV version (eg. 1.11.0), I am told that the bug is fixed so I upgrade to the new version
4. I load in my saved synthesizer from (3) and try to re-train it in order to fix the bug. It doesn't work.

The scenario is confusing because I might imagine re-training a synthesizer will work. But because the synthesizer object was created on an older version, this may lead to incompatibilities.

Expected behavior

Create a new VersionError type and show this if I try to fit on any synthesizer in this scenario.

>>> synthesizer = HMASynthesizer.load('synthesizer.pkl')
>>> synthesizer.fit(data)
VersionError: You are currently on SDV version 1.11.0 but this synthesizer was created on version 1.10.0. Fitting 
this synthesizer again is not supported. Please create a new synthesizer.

Additional context

Apply this error to all synthesizers (available to public and enterprise)

Note that sampling the synthesizer is still ok. (This is what we guarantee.) It's just refitting that is not supported.

Note that the trained version is available under synthesizer.get_info() (the enterprise version may not be -- perhaps we should add this)