fix: hydrate unknown enum values as UNKNOWN and add BANCOMAT card type

[lanz1]

Fixes #127

Problem

The Transactions API returns card.type = "BANCOMAT" for Italian PagoBancomat debit cards. The value is missing from the CardType schema, so the hydrator throws a ValueError and Transactions::get() cannot return a successful transaction at all.

Changes

This PR has two independent commits, so they can be reviewed (or taken) separately:

1. fix(hydrator): when tryFrom() does not recognise a value, Hydrator::castEnumValue() now returns the enum's UNKNOWN case if the enum declares one (CardType, CardResponseType, ReaderStatus, MemberStatus, …). Enums without UNKNOWN keep the current strict behaviour, so testHydrateInvalidBackedEnumValueThrowsValueError is unchanged and still passes. This keeps responses readable when the API adds a value before the SDK is regenerated.
2. feat: add BANCOMAT card type: adds BANCOMAT to components.schemas.CardType in openapi.json and regenerates the four derived enums with the codegen (go -C codegen run . generate ../openapi.json ../src). Running the generator on the unmodified spec produces no diff, so the only changes are the BANCOMAT cases.

I noticed openapi.json is synced from sumup/apis. If the spec change should go there instead, I'm happy to drop the second commit; the first one fixes the crash on its own.

Tests

• testHydrateUnknownValueOfEnumWithUnknownCaseFallsBackToUnknown: an unknown card network hydrates to CardResponseType::UNKNOWN and the other fields are kept.
• testHydrateBancomatCardType: BANCOMAT hydrates to CardResponseType::BANCOMAT.

Locally on PHP 8.5: PHPUnit is green (37 tests) and php-cs-fixer reports no changes. PHPStan reports the same 2 errors in HttpClient/GuzzleClient.php that are already present on main.

[matoous]

Hi! Thanks for the contribution. We generate the SDKs from our OpenAPI specs, so the fix needed to go there (done) and #119 re-generates the SDK from updated specs. I will close your PR in favor of the automatic sync.

Separately, I am taking a look into keeping the enums open (which should have been the behavior in the first place).

Thank you 🙇