docs(tck): the report tallies follow the base's new self-tests

The base gained fourteen tests this pass -- seven for the capability an SDK
cannot express, seven for where the conformance assets came from -- and this
branch runs all of them on top of its own. 276 passed / 41 skipped becomes 289
passed / 42 skipped; the skip is test_spec_assets declining to claim the
submodule pin was checked in an environment that cannot read it.

No scenario count moves. The new pin, 89b1519a, changes two $comment blocks in
canonical-flags.json and no feature file at all, which was checked rather than
assumed: the in-memory suite is 43 passed / 21 skipped / 1 xfailed before and
after, and every one of the thirteen new passes is a unit test.

Measured while doing it, and worth writing down because a later pass will
otherwise re-derive it: results.digest is not a function of the spec pin. Two
runs of the in-memory suite against identical assets produced different digests
and streams 109 bytes apart, because Cucumber Messages carries per-run durations
and ids. Nothing here pins one -- test_report computes the digest it compares,
and the README's example is elided -- so there is no latent flake to fix, and the
claim the README makes for it is the one it can keep: the digest tells a consumer
that what it fetched is what the envelope described, not that two runs asked the
same questions. The stream carries the executed feature source, and that is what
answers the second question.

Signed-off-by: Simon Schrottner <simon.schrottner@flagsmith.com>

Author: aepfli

tools/openfeature-tck/README.md

@@ -950,7 +950,7 @@ the block the schema leaves optional, so an empty one is left out rather than em
 | `test_canonical_set` | the canonical-set guard | a run that executed less than the canonical set fails and publishes nothing |
 
 ```
-276 passed, 41 skipped, 2 xfailed
+289 passed, 42 skipped, 2 xfailed
 ```
 
 No Docker and no network. The conformance suites take under a second; `test_report`,