Use ICU for better Unicode sorting

[alerque]

This is an extract from #1063 with the ICU related bit separately for testing and review.

...and as long as its being run separately here there are probably a few other places that could take advantage of better Unicode support.

Additionally if there is a way (sorting flag?) to do phonetic interspersed sorting instead of per-alphabet that would be an interesting alternative as well.

[simonmichael]

We probably need a little test for this and a mention & example of it in the manual. Also, as previously discussed we are looking for someone on windows (and the other platforms) to test building this, we want to know if it makes installation harder.

[simonmichael]

For the record, this improves the sorting of the new descriptions/payees/notes commands by making them case insensitive and “accent insensitive”, so that eg ã sorts similarly to a. It adds text-icu as a dependency, which requires the icu C library.

[alerque]

The test I added was for a proposed bit that isn't implemented yet (account name sorting), I didn't add a test yet for the part I did implement (descriptions/payee/note sorting) ;-)

[simonmichael]

What's needed to move this forward ? I think:

• find someone to see if this builds on Windows.

And if that works, then:

• clarify which commands are changed, mention the new sorting behaviour in their docs
• consider what other commands to change for consistency
• rebase these changes on latest master, and combine/clean up commits

[alerque]

I think that summary is correct. When the time comes I'll be happy to take care of some of the other steps, but I don't have access to Windows to test on.