Broken handling of unicode f literals

[tlotze]

When applying pyupgrade to code like this: u'foo: {}'.format(1), things break when using tokenize-rt, version 5: the outcome is fu'foo: {1}', which is an invalid literal. Earlier versions work fine and the effect is independent of the pyupgrade version (tested with 3.2.2 and 2.38.2).

[asottile]

garbage in garbage out -- pyupgrade only supports python 3.x code

[asottile]

oh but actually this is still valid 3.x code -- please open an issue on pyupgrade instead -- this isn't a tokenize-rt problem

[asottile]

actually was easy enough to just fix asottile/pyupgrade#762

[tlotze]

Cool, thank you!