fix: Add handling for sub-regions that only consist of "a"

[csandman]

Currently, there is an issue with the inline number replacement in strings consisting of more than just numbers, where the word "a" is always replaced with a "1", even if it's on it's own. This has been mentioned a few times in issues on the original project:

• Why does "a" get converted to 1? finnfiddle/words-to-numbers#36
• Strings converted incorrectly finnfiddle/words-to-numbers#30

This behavior really doesn't make sense, and it's already handled for cases with only the word "a" (in the existing code, the standalone word "a" isn't replaced with anything). So This is meant to fix the rest of the cases where the word has no neighboring numbers to make it into an actual number.

I added some new test cases around this to ensure it works as expected.

// This is an existing test case that already passed
it("a", () => {
  expect(wtn("a")).to.equal("a");
});

// This case also already passed
it("A", () => {
  expect(wtn("A")).to.equal("A");
});

// Previously, these two cases would all result in the "a"s being replaced with "1"
it("A crab cake", () => {
  expect(wtn("A crab cake")).to.equal("A crab cake");
});

it("A cat and a mouse", () => {
  expect(wtn("A cat and a mouse")).to.equal("A cat and a mouse");
});

// This is a valid case where you'd expect the "a" to mean "one hundred"
it("A hundred cats and dogs", () => {
  expect(wtn("A hundred cats and dogs")).to.equal("100 cats and dogs");
});

---

Besides the main change, I made some more cleanup changes to make the code more readable, and organized some files.

[csandman]

This is the change that actually fixed the issue.

[csandman]

This logic was so convoluted. This was basically going through the whole flow, just to finally check if the start index and end index are the same (meaning the string is empty), where as you can just check that at the beginning or filter them out.