Correct what makes a configuration option predicate true

[isHeSatoshi]

Fixes #2004.

The Reference said:

A configuration option. The predicate is true if the option is set, and false if it is unset.

That reads as a single set/unset bit per key, but a name predicate and a key-value predicate for the same key are independent. @epage said in the issue that this needed some further investigation first, so I checked the actual behaviour before writing anything.

Each row below is the --cfg flags passed to rustc on nightly, and the values are cfg!(foo) and cfg!(foo = "bar") in that same program:

--cfg	cfg!(foo)	cfg!(foo = "bar")
(none)	false	false
foo	true	false
foo="bar"	false	true
foo="bar" foo="baz"	false	true
foo foo="bar"	true	true

Two things follow that are worth being precise about:

• A key-value predicate does not satisfy the name predicate, and the name does not satisfy any key-value predicate.
• The last two rows are why the proposed wording "true if the option is set and only a name" would not be accurate on its own: a key can be set as a name and with values at the same time, and then both predicates are true. cfg.option-key-uniqueness already says keys do not need to be unique, which is the same idea one level down.

So the change states the two forms separately and keeps the example inline, since the distinction is easy to get wrong and cheap to check.

cargo run --package=style-check -- src passes. No new links or code blocks are added, so the build and linkcheck jobs are unaffected.