gui: save and load PSBT

[Sjors]

This adds:

• a dialog after Create Unsigned, which lets you save a PSBT file in binary format, e.g. to an SD card
• a "Load PSBT" menu entry lets you pick a PSBT file. We broadcast the transaction if complete

Save flow

By default the file name contains the destination address(es) and amount(s).

We only use the binary format for files, in order to avoid compatibility hell. If we do want to add base64 file format support, we should use a different extension for that (.psbt64?).

Load flow

Select a file:

Offer to send if complete:

Tell user if signatures are missing, offer to copy to clipboard:

Incomplete for another reason:

[DrahtBot]

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Conflicts

Reviewers, this pull request conflicts with the following ones:

• gui: Add a Make unsigned button next to Send #18656 (gui: Add a Make unsigned button next to Send by achow101)
• wip: gui: Drop connectSlotsByName usage #18246 (wip: gui: Drop connectSlotsByName usage by promag)
• "PSBT Operations" dialog #18027 ("PSBT Operations" dialog by gwillen)
• Bugfix: GUI: Restore SendConfirmationDialog button default to "Yes" #17463 (Bugfix: GUI: Restore SendConfirmationDialog button default to "Yes" by luke-jr)
• gui: Fix manual coin control with multiple wallets loaded #17457 (gui: Fix manual coin control with multiple wallets loaded by promag)
• [BIP 174] PSBT version, proprietary, and xpub fields #17034 (Bip174 extensions by achow101)
• [BIP 174] Implement serialization support for GLOBAL_XPUB field. #16463 ([BIP 174] Implement serialization support for GLOBAL_XPUB field. by achow101)
• Try to use posix_fadvise with CBufferedFile #14485 (Try to use posix_fadvise with CBufferedFile by luke-jr)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

[practicalswift]

I haven't investigated thoroughly yet but it seems like at least a subset of the issues regarding handling of malformed PSBT inputs reported in #17149 ("Potential PSBT issues found by the PSBT fuzzer") are reachable from this code?

That is not a problem once the #17149 issues have been fixed by making the PSBT code more robust against malformed PSBT input, but I wanted to mention the dependency here so that the PSBT issues get fixed before the merge of this PR :)

[Sjors]

@achow101 I uncommented this because the next role was UPDATER, which seems similar to what @justinmoon found: https://bitcoin.stackexchange.com/questions/89203/analyzepsbt-says-next-role-is-updater-but-nothing-is-missing

Will investigate more thoroughly when I work on this PR again, if I still see it.

[gwillen]

UI comment: I think "create unsigned" / "save" is a weird juxtaposition. Perhaps "Copy PSBT" / "Save PSBT"? Or perhaps "Create Unsigned" with further options in a dialog?

[promag]

Concept ACK.

[Sjors]

Rebased. I can address feedback for #17463 here if needed.

Or perhaps "Create Unsigned" with further options in a dialog?

Yes, this seems like a better idea. It's especially useful to have some flexibility after loading a PSBT.

[Sjors]

I added a dialog for Create Unsigned that appears after the usual transaction approval. The PSBT is automatically copied to clipboard as before, but the dialog offers to also save it. When loading a PSBT, the dialog asks if the users wants to broadcast. When there's a problem, e.g. missing signatures, it offers to copy it (so you can e.g. inspect it with RPC).

[Sjors]

I dropped the dependency on #17463; should be easy to rebase. This is now ready for review.

[gwillen]

Tested ACK e6ea593. Thoughts:

• Obviously I'm not exactly an unbiased reviewer since some of this is based on some of my work, but I think my ACK probably still counts. :-)
• I don't like the verbosity of the save filenames, but that's so easy to change later, and so subject to bikeshedding, that I think it should be left alone for this PR.

[luke-jr]

No way to actually sign an incomplete one?

Maybe it should load raw signed transactions too...

[Sjors]

@luke-jr both should be possible, but I prefer to keep this PR simple.

@practicalswift wrote:

I haven't investigated thoroughly yet but it seems like at least a subset of the issues regarding handling of malformed PSBT inputs reported in #17149 ("Potential PSBT issues found by the PSBT fuzzer") are reachable from this code?

That is not a problem once the #17149 issues have been fixed by making the PSBT code more robust against malformed PSBT input, but I wanted to mention the dependency here so that the PSBT issues get fixed before the merge of this PR :)

These are merged now in #17156, though I haven't rebased to preserve @gwillen's ACK.

[luke-jr]

Typo

[Sjors]

@luke-jr that's already fixed in master