Global xpubs are stored in a map of key origin to set of xpubs, while the serialization writes one record per xpub, keyed by the xpub. Merge unions the map origin-by-origin, so when the combined PSBTs provide different key origins for the same xpub, the result serializes the same PSBT_GLOBAL_XPUB key twice. BIP 174 declares PSBTs with duplicate keys invalid and the deserializer rejects them, so combinepsbt returns a PSBT that no RPC can parse again. This affects all releases since the merge loop was added in #17034 (v23.0).
<details><summary>Reproduction on master</summary>
The PSBTs share the unsigned transaction and xpub, and differ only in the master fingerprint of the global xpub record (00000000 vs 11111111):
$ A=cHNidP8BADwCAAAAAaqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqAAAAAAD/////AQAAAAAAAAAAAAAAAABPAQQ1h88AAAAAAAAAAACHPf+BwC9SViP9H+UWfqw6VaBJ3j0xS7Qu4if/7TfVCAM5o2ATMBWX2u9B++WToCzFE9C1VSfsLfEFDi6P9JyFwgQAAAAAAAAA
$ B=cHNidP8BADwCAAAAAaqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqAAAAAAD/////AQAAAAAAAAAAAAAAAABPAQQ1h88AAAAAAAAAAACHPf+BwC9SViP9H+UWfqw6VaBJ3j0xS7Qu4if/7TfVCAM5o2ATMBWX2u9B++WToCzFE9C1VSfsLfEFDi6P9JyFwgQRERERAAAA
$ bitcoin-cli -regtest decodepsbt "$(bitcoin-cli -regtest combinepsbt "[\"$A\",\"$B\"]")"
error code: -22
error message:
TX decode failed Duplicate Key, global key "01043587cf00...9c85c2" already provided: iostream error
</details>
Deduplicate by xpub when merging, keeping the origin that is already present: BIP 174 lets the Combiner "pick arbitrarily when conflicts occur", and conflicting unknown and proprietary records are already resolved the same way. The logic is shared between combinepsbt and joinpsbts through a new MergeGlobalXPubs helper. The second commit adds a test that fails on master with the error above, and the last commit removes the global_xpubs tracking set in Unserialize, write-only since the generic duplicate key check introduced in #21283 (1e2d146b47) replaced the explicit one.
Note: the xpub loop in joinpsbts currently has no observable effect, since the collected xpubs never reach the returned PSBT. My #35516 fixes that, so this PR should land first: on its own, #35516 would make the same duplicate key issue reachable through joinpsbts, while with the shared helper in place it never becomes reachable. I will rebase #35516 on top afterwards.
