[MAINTENANCE] Remove the follow_imports = 'silent' override block from the mypy configuration

Part of #12188.

What to do

Delete the [[tool.mypy.overrides]] block in pyproject.toml that sets follow_imports = 'silent' for ten module patterns — the one annotated "TODO: remove these overrides once we have typed the modules" — and regenerate the relaxation inventory in the same change.

Once #12190 and #12191 have merged, every great_expectations.* and tests.* entry in that block is inert: those modules are in files, so mypy checks them as build targets and follow_imports no longer governs them. The one entry that can still do work is contrib.experimental.great_expectations_experimental.*, because contrib/ is deliberately outside files and is reached only by import. Done means: the block is gone, the run is green, and contrib/ still reports nothing — by one explicit, documented decision rather than a leftover.

Where

• pyproject.toml — delete the whole silent-imports override block. Then, depending on what you measure (below), possibly add this one block in its place, comment included:

[[tool.mypy.overrides]]
# Community-contributed modules are maintained outside the checked package and are
# deliberately not part of the enforced type-check; their errors are not reported.
module = "contrib.experimental.great_expectations_experimental.*"
ignore_errors = true

• scripts/mypy_relaxation_inventory.json — regenerate with the guard's emit mode, never by hand. The ten silent entries drop out; the ignore_errors entry, if added, appears.
• Nothing under contrib/ changes.

How to verify

! grep -q "follow_imports = 'silent'" pyproject.toml && invoke type-check --ci

Then measure once more with the 'tests/expectations', exclude line also removed locally (do not commit that removal — it belongs to #12131), because four modules under tests/expectations/core/ import great_expectations_experimental and that is the only edge that still reaches contrib/: regenerate the inventory, run invoke type-check --ci again, and count lines beginning with contrib/.

Before: the fence exits 1 because the block is present; 945 source files checked, zero errors (at 3798c021b with #12189–#12191 applied). Deleting the block alone: still zero. Deleting it with tests/expectations also lifted: errors under contrib/ appear (25 at the last measurement that reached them). After: the fence exits 0; zero errors in both measurements; pytest tests/expectations/core unchanged.

Requirements

1. When the change merges, no [[tool.mypy.overrides]] entry sets follow_imports = 'silent'.
2. The type-check reports zero errors and no error whose path begins with contrib/ — both as-is and with the tests/expectations exclusion lifted locally.
3. If the second measurement reports contrib/ errors, the single ignore_errors override above is added in this change, verbatim, comment included. No error in contrib/ is fixed, no file under contrib/ changes, and no silent-imports entry is retained.
4. If either measurement reports a new error outside contrib/, comment on this issue with the listing before proceeding — that is a different problem than this issue describes.
5. The inventory must be regenerated by python scripts/mypy_config_guard.py --emit-inventory in the same change, and the guard must pass.
6. pytest tests/expectations/core tests/render must pass with the same outcomes as before.

Out of scope: the tests/expectations exclusion itself (#12131); any exclude pattern (#12190, #12191, #12128); any strictness flag; any edit under contrib/.

Order

• Blocked by #12191 — while any library exclusion stands, the block is doing real work: deleting it before the lifts reveals 481 errors, after them at most 25, all in contrib/.

Notes

• The order between this issue and #12131 is not fixed, which is why the second measurement exists. If #12131 merges first, the plain deletion reveals the contrib/ errors directly and the override is clearly needed. If this merges first without the override, #12131 inherits those errors on a change that has nothing to do with contrib/. Measuring both ways and adding the override when either shows contrib/ errors keeps that cost off the other issue.
• Do not "fix" the cost by removing the great_expectations_experimental imports from those four test modules — that changes what the tests exercise and is #12131's surface anyway.