Write the documentation an operator reads before installing the first release

[iderex]

The README today describes an intention. At the first release it has to describe
a thing that exists, and it has to be honest about how small that thing is,
because the alternative is somebody installing it expecting a client.

What this produces

A first paragraph saying what this is and what it is not, in the register the
organisation's own state table already uses.

What an operator needs to run the probe, what it will tell them, and what to do
with the output.

The data page from #74 linked and summarised, since somebody deciding whether to
point this at their own server is deciding what happens to their data.

The licence from #12 named.

The known limitations, listed rather than implied, including every published
speed number this release does not yet measure.

Done when

README.md describes the released thing rather than the intention, every claim
in it is either checkable by a command it names or marked as a claim, and a
person who has never seen the repository can install and run the probe from it
alone. Blocked on #12, #74, #91, #92.

[iderex]

A closed issue left a publication condition unmet and named this document as one
of the two places it is met. The list above has a line that looks like the same
obligation and asks for something else. Read at
e4903a96a909ea5d2a1300b9a04100e3a461911a.

The line here:

gh issue view 95 --json body --jq '.body' | tr '\n' ' ' | tr -s ' ' | grep -o 'The known limitations[^.]*\.'
The known limitations, listed rather than implied, including every published speed number this release does not yet measure.

What docs/decisions/0064-the-numbers-the-core-does-not-report.md is waiting on
this document for:

sed -n '114,119p' docs/decisions/0064-the-numbers-the-core-does-not-report.md
## Where the statement is published, and why not yet

#64 asks for the statement in the documentation an operator and a client author
both read. Neither document exists. #95 is the documentation an operator reads
before installing, #75 is the client contract, and both are open with nothing
landed.

The two are not the same sentence. A number this release does not yet measure is
a number that arrives later. The two numbers that record is about are ones the
core has no interval for and will not report, and the difference is the whole
point of it:

sed -n '100,105p' docs/decisions/0064-the-numbers-the-core-does-not-report.md
An empty span field rather than a second shape, because a client reporting four
numbers through two shapes writes the second one badly, and because the empty
field is the statement this record exists to make, arriving where somebody
reading the data will meet it. A reader holding a report for focus change can see
that no core interval was claimed, without having read 0008.

An operator who reads the line above as written puts both kinds under one heading
and takes the whole list as temporary, which is the reading the record was written
to prevent. So this document carries two statements rather than one, and the
second one is not a limitation of the release.

Two smaller things the same record fixes about how the sentence is written here.

It is a reference rather than a restatement:

sed -n '169,173p' docs/decisions/0064-the-numbers-the-core-does-not-report.md
the one document that exists and that both readers reach, so the condition would
be met in the world rather than on paper. It costs a fourth hand in a file three
other issues are written against, and it costs a summary that will disagree with
#95 and #75 on the day either lands, which is the drift this repository counts as
a defect class rather than an accident.

And where the sentence does not fit the reader it is for, the record gives way
rather than this document:

sed -n '204,207p' docs/decisions/0064-the-numbers-the-core-does-not-report.md
#95 or #75 lands and its author finds the sentence here does not fit what its
reader needs. The record that replaces this one is written from that document
rather than the document being written from this record, since the reader it
failed is the one it was for.

I have changed no condition and no scope. The three conditions here are untouched.