chore: delete two pieces of dead configuration

[lsimons]

Two unrelated dead-config removals, one commit each so either can be dropped.

1. gitlab_group on MappingEntry (#42)

src/mdd/sharepoint/mapping.py carried a GitLab-specific default —
gitlab_group: str = field(default="mdd/sharepoint") — into the provider-neutral
core. Two independent problems: it is a house default for a specific group path
on a specific GitLab instance, sitting in the layer whose mirror seam exists so
that a backend supplies exactly that; and the field was dead — parsed,
validated, defaulted, stored, and never read.

The issue could not determine whether a downstream wrapper read it off
MappingEntry. That has now been checked: neither the private SBP wrapper nor
the public lsimons/mdd-wrapper reference wrapper contains any reference to
gitlab_group. So it is simply dead, and this deletes it — field, constant,
parse branch, the config-example lines in mapping.py and
configs/sharepoint-mapping.yaml, the documented key in S10, and the one test
that only asserted the parse round-trip.

Fixes #42

2. The test-corpus mise task (#48)

mise run test-corpus has never worked in this repository. It pointed at
tests/confluence/test_corpus_smoke.py, which is absent from the working tree
and from all of git history — the task was already dangling in the initial
commit. pytest exits 4 (usage error), so it was a hard task failure, not an
empty-but-passing run. It went unnoticed because test-corpus is not in the
ci task's depends list.

There is no Confluence corpus smoke test anywhere, and the corpus is already
exercised by mise run ir-coverage, mise run ir-roundtrip and
tests/sharepoint/test_corpus_matrix.py, each with its own task. So the task is
deleted rather than repointed at a test that already has one — which also
removes a misleading entry from mise tasks.

Fixes #48

Not done here

The issue raises whether a check that every [tasks.*].run path exists would be
worthwhile, since a dangling task path is silent today. Deliberately out of
scope; worth a follow-up issue.

Co-Authored-By: lsimons-bot bot@leosimons.com