levenshteinDistance does not support unicode characters

Details:

levenshteinDistance does not support unicode characters outside of the BMP range properly.
Step to reproduce: expect(levenshteinDistance('\u{1f431}', '')).toBe(1) // got 2

Fix

I can issue a PR with the fix https://github.com/dubzzz/javascript-algorithms/commit/856ba360134956cca2bed165929a4d4409537201

How did I find it?

Thanks to property based testing framework fast-check.
The property was the following:

import fc from 'fast-check';

fc.assert(
    fc.property(
      fc.fullUnicodeString(), fc.fullUnicodeString(),
      fc.fullUnicodeString(), fc.fullUnicodeString(),
      fc.fullUnicodeString(),
      (aBegin, aEnd, bBegin, bEnd, common) =>
        levenshteinDistance(aBegin + common + aEnd, bBegin + common + bEnd)
          <= Math.max([...aBegin].length, [...bBegin].length) + Math.max([...aEnd].length, [...bEnd].length)
    )
  )

Or:

for any a and b - strings
such that a = aBegin + common + aEnd and b = bBegin + common + bEnd
levenshteinDistance(a, b) is at most max(numChars(aBegin), numChars(bBegin)) + max(numChars(aEnd), numChars(bEnd))

Start by locating the levenshteinDistance implementation and its existing tests. Reproduce the astral Unicode case from the issue, then add regression coverage for it and verify that the distance for '\u{1f431}' versus an empty string is 1.