TreeSpellChecker: cases where it doesn't work as expected?

I've just started playing with TreeSpellChecker and found a couple of cases where it doesn't give me the corrections I would expect.

If I run this script in ruby-3.0.3:

#!/usr/bin/env ruby
require 'set'

paths = [
  "dir/typo",
  "typo",
]
sc = DidYouMean::TreeSpellChecker.new(dictionary: paths)
puts "*** none slashed:"
puts sc.correct("tyop")

paths = [
  "dir/typo",
  "typo",
]
sc = DidYouMean::TreeSpellChecker.new(dictionary: paths)
puts "*** query slashed:"
puts sc.correct("/tyop")

paths = [
  "dir/typo",
  "/typo",
]
sc = DidYouMean::TreeSpellChecker.new(dictionary: paths)
puts "*** dict slashed:"
puts sc.correct("tyop")

paths = [
  "dir/typo",
  "/typo",
]
sc = DidYouMean::TreeSpellChecker.new(dictionary: paths)
puts "*** both slashed:"
puts sc.correct("/tyop")

I get this output:

*** none slashed:
*** query slashed:
dir/typo
*** dict slashed:
*** both slashed:
/typo

And I think in all the cases, the second path on the list (either typo or /typo) should be suggested.

What do you think? ✨✨✨
cc @yuki24

Start with the TreeSpellChecker entry point and run the four Ruby examples from the issue to reproduce the slash and dictionary combinations. Trace how correct handles these paths, then verify that each case suggests the second path shown in the examples.