Add opt-in SEPA Instant (SCT Inst / Echtzeitüberweisung) support to credit transfers

[wolframarnold]

What

Adds opt-in SEPA Instant Credit Transfer (SCT Inst / Echtzeitüberweisung) support to SepaTransfer. When the config carries "instant": True, the generated pain.001 emits PmtTpInf/LclInstrm/Cd = INST — the EPC/DK mechanism for requesting an instant transfer — positioned directly after SvcLvl, per the ISO 20022 PaymentTypeInformation sequence (InstrPrty?, SvcLvl?, LclInstrm?, CtgyPurp?).

sepa = SepaTransfer({..., "instant": True}, schema="pain.001.001.03")

<PmtTpInf>
  <SvcLvl><Cd>SEPA</Cd></SvcLvl>
  <LclInstrm><Cd>INST</Cd></LclInstrm>
</PmtTpInf>

Why

Under the EU Instant Payments Regulation (2024/886), euro-area PSPs must offer instant credit transfers at price parity with standard ones, so requesting INST is becoming a routine need. sepaxml already emits LclInstrm/Cd for direct debits but had no way to do so for credit transfers. (Downstream, e.g. python-fints pins sepaxml==2.7.* for its SEPA payloads.)

Design

• Config-level flag, not per-payment. PmtTpInf is a payment-information-block element shared by all grouped transfers; a per-payment flag on a block-level node is ambiguous for mixed batches. (This is the flaw visible in prior-art Add structured reference and instant payment support #71's per-payment approach.)
• Mirrors the library's own direct-debit precedent. debit.py already sets LclInstrm/Cd from self._config['instrument'] (CORE/B2B); this is the symmetric credit-transfer counterpart.
• Scoped to SEPA. The INST node is nested inside the existing non-domestic guard, since INST is a SEPA local instrument — a domestic transfer has no SvcLvl=SEPA/PmtTpInf block to attach it to.

Tests

tests/transfer/test_instant.py asserts INST is present directly after SvcLvl (batch and non-batch paths), XSD-valid, and absent without the flag. Verified XSD-valid across all four supported pain.001 schemas (03/09/10/11). No existing tests change.

Real-world verification

The exact XML this change produces (<PmtTpInf><SvcLvl><Cd>SEPA</Cd></SvcLvl><LclInstrm><Cd>INST</Cd></LclInstrm></PmtTpInf> on a pain.001.001.03 batch, batch=True) was verified end-to-end against a live German bank on 2026-07-16, using two €0.01 transfers between real accounts at two different institutions:

• Upload & acceptance: Nassauische Sparkasse accepted the INST-flagged file through its bulk-order file-upload channel (Sammelaufträge → Datei-Übergabe); the file was parsed completely (message ID, batch-booking flag, item count, amount, recipient), passed the Verification-of-Payee check at release, and was executed after a single TAN.
• A/B control: an otherwise-identical transfer without LclInstrm was sent the same day. The receiving bank (Taunus Sparkasse) booked the flagged transfer as ECHTZEIT-GUTSCHRIFT (SEPA Instant credit) and the unflagged one as ONLINE-UEBERWEISUNG (standard SCT) — confirming the element alone switches the scheme, cross-bank, and is not silently downgraded.
• Both transfers booked same-day; the instant one posted within seconds.

(The verification was done by injecting this element into sepaxml 2.7.0 output and re-validating against the bundled pain.001.001.03.xsd — i.e., precisely the XML shape this PR generates natively.)

Prior art

#71 established the same element and position (good corroboration), but bundled it with a per-payment API, structured references, six new XSDs, and a whole-file reformat (~9.4k additions across 24 files) and was self-closed unreviewed. This PR is the minimal form: one feature, ~13 lines in transfer.py, one new test file, one README line — no reformatting, no new schemas.