docs: the whole documentation checked against the code and written for a newcomer

[rducom]

Intent

Make the documentation something a newcomer can rely on and read with pleasure. An audit of every page against the code found a few dozen statements that had drifted (stale outputs, behaviours that changed, members that are internal), and the prose read as working notes: dashes, semicolons, section numbers as cross-references, internal step codes, and "few words. Then the sentence" openings.

The branch fixes the facts and rewrites the prose of the guide, the design documents, the READMEs and the contribution notes. Each page now says the essential first, in whole sentences, at an engineer's level.

What it brings

• The guide's 50 pages match the library today, their sample outputs taken from a fresh run of the samples and of vxdump.
• The 19 design documents link sections by name instead of by number. The numbered headings stay, so existing anchors still resolve.
• The errors page documents the memory, torn-commit and integrity exceptions, and the observability page documents the event counters and the grouping statistics, which were missing.
• Three XML summaries and three code comments no longer state wrong behaviour or cite plans that are not in the repository.

Among the corrections: a chunk of one value is written as a constant, groups come in no promised order without an OrderBy, KeyCountAsync counts the entries under the current key, owned batches keep their selection, lists of records and maps are supported, a decimal bucket is not implemented, the corpus is 876 files and the per-encoding corpus 61, and vxdump exits with 2 on an unknown option.

Risks and tests

• Documentation, comments and two sample labels only. No behaviour changes.
• dotnet build Vorticity.slnx -c Release passes with no warning, so every <see cref> resolves.
• Every relative link and anchor in the rewritten pages was checked.
• The test suite was not run, since no code path changed.

Left as it is, on purpose

• The generated sections of the benchmark pages still hold semicolons. Their text comes from the bench's own strings and would be overwritten by the next run, so they change in the generator, not here.
• The caveats and skipped entries of the corpus manifest still say vortex-row does not exist. They are written by conformance-gen and change with a corpus regeneration.
• Exception messages quoted in the guide keep their own punctuation, since they are the program's output.
• The crosscheck figure in the bench README (854 files) was not measured again.

Commits

1. The guide, with the two sample labels that follow the code.
2. The design documents.
3. The READMEs, contribution and security notes, the bench and tools pages.
4. XML summaries and comments.