Port Swift integration tests to pytest.

[criemen]

Tests pass on this PR due to a test infrastructure failure, they ought to fail.

Apparently some test databases have DB-check failures, instead of blanket-disabling db-checks for Swift, I've marked these tests now manually.

[criemen]

I'm seeing the following test failure:

ql/swift/ql/integration-tests/posix-only/deduplication/test.py::test::BuiltinTypes.ql FAILED - RESULT
------------------------------ Captured diff (QL) ------------------------------
BuiltinTypes.expected <-> BuiltinTypes.actual
--- expected
+++ actual
@@ -2,7 +2,6 @@
 | Builtin.Executor | BuiltinExecutorType |
 | Builtin.FPIEEE32 | BuiltinFloatType |
 | Builtin.FPIEEE64 | BuiltinFloatType |
-| Builtin.FPIEEE80 | BuiltinFloatType |
 | Builtin.Int1 | BuiltinIntegerType |
 | Builtin.Int8 | BuiltinIntegerType |
 | Builtin.Int16 | BuiltinIntegerType |

I have no idea why this triggers on the pytest port, but not on the original branch.

[redsun82]

I'm seeing the following test failure:

ql/swift/ql/integration-tests/posix-only/deduplication/test.py::test::BuiltinTypes.ql FAILED - RESULT
------------------------------ Captured diff (QL) ------------------------------
BuiltinTypes.expected <-> BuiltinTypes.actual
--- expected
+++ actual
@@ -2,7 +2,6 @@
 | Builtin.Executor | BuiltinExecutorType |
 | Builtin.FPIEEE32 | BuiltinFloatType |
 | Builtin.FPIEEE64 | BuiltinFloatType |
-| Builtin.FPIEEE80 | BuiltinFloatType |
 | Builtin.Int1 | BuiltinIntegerType |
 | Builtin.Int8 | BuiltinIntegerType |
 | Builtin.Int16 | BuiltinIntegerType |

I have no idea why this triggers on the pytest port, but not on the original branch.

I think those tests are unfortunately sensible to what runner runs them, so it might just be a question of having changed that moving to the integration test check template.

We can just accept the change, or put some effort into making the test a bit less host dependent, like skipping some of the entries in the query. Unfortunately testing Builtin stuff is inherently host-dependent...

[criemen]

The query already skips Builtin.FPIEEE16, so I'd be happy to also skip Builtin.FPIEEE80 there too. I'll check if we moved from intel to arm macs here, as Builtin.FPIEEE80 sounds very much like an x86_64 thing, it'd make sense if it disappears there.