README: five sentences fixed for the documents gate (closes #229)

The documents gate on `next` failed five README sentences
(#219 (comment)). Each
is replaced with the gate's rewrite after a check against the code: the
keywords score sentence is split in two; `-1` is dropped from the
invalid ID examples, since the option parser refuses a positional `-1`
as an unknown option before the ID check; `mlData` drops the whole
`clip` block, as `src/backup.ts` and `src/metadata-backup.ts` do; a
collection name gets its ID at 250 bytes and a file name at 255, as
`uniqueNames` does; `lastTriedAt` is one timestamp per run. Nothing
else changes; `make fmt` reflowed the touched paragraphs.

Model: fable-5-1

README.md

@@ -472,8 +472,8 @@ fresh reads"). `backup` counts it as failed (see "What quak cannot decrypt").
 
 A collection ID or file ID given to `files`, `get`, `get-thumb` or
 `fix-missing-thumbnails --file` is written in decimal digits and is at
-most 9007199254740991. Anything else, such as `0x10`, `1e3` or `-1`, is refused
+most 9007199254740991. Anything else, such as `0x10` or `1e3`, is refused before
-before any request is sent. The command logs `Invalid file ID: <arg>` or
+any request is sent. The command logs `Invalid file ID: <arg>` or
 `Invalid collection ID: <arg>` and exits 1.
 
 ### What quak prints
@@ -692,10 +692,10 @@ Update times are in microseconds, as Ente records them.
 The JSON also holds the machine-learning data Ente computed for the photo, as
 `mlData`. Ente's apps detect faces in each photo and compute a CLIP embedding
 for it, a list of 512 numbers that describes what the photo shows. `mlData`
-holds everything but that embedding. In its place the JSON holds `keywords`,
+holds everything but the `clip` block, which holds that embedding. In its place
-derived from the embedding as described under "Keywords". A file with no
+the JSON holds `keywords`, derived from the embedding as described under
-machine-learning data has neither field. A file whose data holds no usable
+"Keywords". A file with no machine-learning data has neither field. A file whose
-embedding has no `keywords`.
+data holds no usable embedding has no `keywords`.
 
 The backup waits for the library's fetch of machine-learning data to finish
 before it writes the JSON files. When a file's data cannot be fetched,
@@ -748,11 +748,11 @@ derives a file's keywords the way the app sorts it. The categories are:
 
 Each category has a query, and a text embedding of that query from Ente's CLIP
 text model. A file's score for a category is the cosine similarity of the file's
-embedding and the category's: a number from -1 to 1, higher the more alike the
+embedding and the category's. That is a number from -1 to 1; the more alike the
-two embeddings are. The file falls in the category when its score reaches the
+two embeddings are, the higher it is. The file falls in the category when its
-category's minimum score. The keywords are the titles of the categories the file
+score reaches the category's minimum score. The keywords are the titles of the
-falls in, highest score first. A video never falls in a category marked "no
+categories the file falls in, highest score first. A video never falls in a
-videos", as in the app.
+category marked "no videos", as in the app.
 
 #### Collection directories
 
@@ -775,11 +775,11 @@ ignoring case, each gets its ID added: two albums named `Trip` become
 `Trip (10)/` and `Trip (11)/`. Two files in one collection whose titles match,
 ignoring case, get their IDs added before the extension: `IMG_0001 (12345).JPG`
 and `IMG_0001 (12346).JPG`. A collection named like another collection's JSON
-file, such as `Trip.json` beside `Trip`, also gets its ID added. A name too long
+file, such as `Trip.json` beside `Trip`, also gets its ID added. A file name
-for a file name, which most file systems limit to 255 bytes, gets its ID added
+longer than 255 bytes, the limit on most file systems, gets its ID added too. So
-too. It is then cut short before the ID, just far enough to fit. IDs never
+does a collection name longer than 250 bytes, which leaves room for `.json`. It
-change, so a name stays the same from run to run until such a clash appears or
+is then cut short before the ID, just far enough to fit. IDs never change, so a
-goes away.
+name stays the same from run to run until such a clash appears or goes away.
 
 Each run removes every symlink in a collection's directory that points into the
 date folders and is not one of its current links, yours included. A file or
@@ -834,7 +834,8 @@ Each entry holds:
 - `fileID` and `title`;
 - `classification`: the kind of failure;
 - `attempts`: how many runs in a row have failed the file;
-- `lastTriedAt`: when it was last tried, as a Unix time in milliseconds;
+- `lastTriedAt`: the start of the run that last tried it, as a Unix time in
+  milliseconds;
 - `error`: the error, as quak printed it.
 
 The kinds are: