Rails serialises `BigDecimal`s as JSON strings to prevent loss of
precision. The `latitude` and `longitude` columns in `StopArea` are
stored as `BigDecimal`s. The trouble is that TomTom's API requires the
latitude & longitude values to be JSON floats, not strings.

Make a new JSON serialiser that converts the `BigDecimal` coordinates to
float to allow the values to be correctly interpreted by the API.

Refs #6222

Using a set ended up not working out. I needed to be able to index into
the list in `#extract_costs_to_way_costs!`, and sets aren't indexable.
This is because they're supposed to be unordered, though modern Ruby
implements `Set` with `Hash` under the hood, which is ordered in Ruby.

I like the idea of having a data structure that automatically eliminates
duplicates, but it wasn't meant to be, because for the extraction to
`WayCost`s, I need an ordered list. Rather than create a new
`OrderedSet` type, I just went the simple route and used an Array,
eliminating the duplicates manually because I know when duplicates are
supposed to occur due to the nature of the data set.

Remove the `#eql?` and `#hash` methods from `TomTom::Matrix::Point`.
Because we're not longer using `Set`, these methods don't need to be
implemented.

Refs #6222

We need to persist stop IDs in order to properly construct `WayCost`
objects from the costs returned from the TomTom matrix API.

In order to persist stop IDs, my idea here is to group together a point
and its corresponding ID into a bucket. When we later
`#extract_costs_to_way_costs!`, we'll be able to grab the correct ID for
a given coordinate to create a `WayCost` from it.

Here, we create a new `TomTom::Matrix::Point` class that encapsulates a
coordinate and an ID, and build a `Set` of those. I needed an `#eql?`
and `#hash` method on `Point` as described in the `Set` documentation
(https://ruby-doc.org/stdlib-1.9.3/libdoc/set/rdoc/Set.html) in order to
properly maintain a unique set.

Refs #6222