trezor: support trezorlib==0.13.9, add compat for <=0.13.8. fixes #9116

[accumulator]

No description provided.

[ecdsa]

Do we we really want to support incompatible versions of trezorlib?
Testing the trezorlib version by try.. except ImportError is ugly. If we keep support for the old version, we should be checking trezorlib.__version__ instead.

I am not saying we should do that, though. I would rather drop support for the old version, unless we have a good reason to keep supporting it. Has this been thought through?

[accumulator]

I prefer dropping support for the old version, however I haven't had time to test against 0.13.9 so this PR is just meant as a convenience to test against 0.13.9

[PiRK]

On my end the reasoning was that 0.13.8 is well tested and I would rather use it for another release or two while people running from source can start testing 0.13.9 in real life scenarios and confirm that there is no other hidden broken compatibility issue.

[dlitz]

I just noticed this after submitting my PR, but if you want a simpler change that only supports 0.13.8 and 0.13.9, I made a patch that does that in #9141.

[SomberNight]

I have merged the other PR (#9141).
That still has the "compat" part but in a cleaner way IMO.
Thanks for the comments and code itself in both PRs.

Do we we really want to support incompatible versions of trezorlib?
I would rather drop support for the old version, unless we have a good reason to keep supporting it.

In general we don't, I agree, but in this case it can be done relatively cleanly.

Do we we really want to support incompatible versions of trezorlib?
Has this been thought through?

One reason to support somewhat older python-trezor, would be that debian is backporting new versions of electrum much more frequently than new versions of python-trezor. Atm debian stable (backports) has electrum 4.5.5 but only python-trezor 0.12.4. i.e. they are incompatible (debian ticket). Anyway, clearly the gap is too large and this specific change does not help.

Still, in general I don't think it is worth a lot of code complexity or effort.