This is more consistent with the test layout in the `isFilesPattern`
suite, and elsewhere in this file.

Personal taste, but this makes it easier to see that the pattern is
identical in each pair of lines, and we've tended to use `.len`
elsewhere.

Before:
  no placeholder
  with placeholder
  multiple placeholders

After:
  no placeholder
  one placeholder
  multiple placeholders

Previously, an IndexDefect was raised for an input string that ends with
the % character.

Let's put both these variables close to the `isString` proc that uses
them.

Appease `nimpretty`.

I think this reads more naturally, and we've tended to use this style
elsewhere.

This makes it obvious that `filesPatterns` isn't used elsewhere.

This makes the name of this variable more consistent with the code in
`isFilesPattern`:

    for ph in extractPlaceholders(s):
      if ph notin filesPatterns:
        result = false

The Nim compiler doesn't optimize the previous pattern.

The recommended ways to write it are these:

    foo.add bar
    foo &= bar

This makes it obvious that we don't mutate `c`.

This is more consistent with the name of the parameter in
`isFilesPattern`, and our code elsewhere.

This was used neither in the configlet codebase, nor the tests.

Probably more readable.

I don't have a strong preference for this indentation style, but it's
what we've used elsewhere.

It was a little confusing that passing `isRequired = false` to
`hasObject` did nothing, because we checked `hasKey` twice.

But the problem is that the existing `hasX` design is a bit of a mess,
or at least the proc names. I don't love that a proc named `hasObject`
returns true when `isRequired = false` is passed.

Let's just do it this way for now - I found it little clearer.
Eventually, all this `hasX` code might be refactored.

This makes it more obvious that this func was only used during the
tests.

After the Nim 1.6 release we'll be able to test procs in other modules
without exporting them.

This test case used 3 placeholder characters, not 2.

This reverts commit 6df6fe6.

Oops. It does need to be like this. Otherwise, when the optional `files`
key is missing, `configlet lint` incorrectly prints this:

    The `files` key is missing:
    ./config.json

This is a very basic check, but at least it catches something like

    %{snake_slug.foo

Currently, there is only one track with multiple placeholders.

And the (unlikely) use of 3 placeholders.

We omitted these elsewhere.

Note that the string is still set at compile-time.

The func is a bit long - oh well.

This better communicates that `extractPlaceholders` isn't printing error
messages, so we don't need to iterate over every placeholder.

This makes the code use the terminology from the docs:

4. Valid `files` pattern: A non-blank string¹ that specifies a location
   of a file used in an exercise, relative to the exercise's directory.
   A pattern may use one of the following placeholders:
   - `%{kebab_slug}`: the `kebab-case` exercise slug (e.g. `bit-manipulation`)
   - `%{snake_slug}`: the `snake_case` exercise slug (e.g. `bit_manipulation`)
   - `%{camel_slug}`: the `camelCase` exercise slug (e.g. `bitManipulation`)
   - `%{pascal_slug}`: the `PascalCase` exercise slug (e.g. `BitManipulation`)

"valid files pattern" sounds like it could be missing a possessive
apostrophe.

Let's refer to a single string as either a
  `files` pattern
or a
  file pattern

The latter reads better here, since the key name is already mentioned.

This is allowed, but unlikely.