MID stability issue with format=flowed tests

[sebb <se...@gmail.com>]

The unit tests agree with each other for the current code and for
versions 0.11 and 0.10.
Not all tests pass, but the same tests fail in the same ways for each version.
This is good, as it shows that the generators are consistent (at least
for the existing test data).

However these versions do not always agree with the values generated
by the 'foal' version.
All the discrepancies relate to format=flowed messages, but it does
not affect all such messages.

This is because the code for handling such messages was broken, such
that the conversion was never done. This has now been fixed in the
'foal' version.

Whilst it may be good to fix the bug, unfortunately it changes the
generated MIDs.

Furthermore, since the output depends on how the convertToWrapped code
behaves, any fixes to its code will alter the MIDs further.

Rather than create yet more MID variants, I think it would be sensible
not to apply the conversion to messages *unless* using a generator
that depends on the raw body only.

Alternatively, perhaps it would be possible to apply the conversion
only *after* MID generation.

Sebb.