Add IO overloads to Char#upcase, #downcase, #titlecase

[HertzDevil]

These additions mirror String#upcase, #downcase, and #capitalize respectively, which also have overloads that accept an IO.

There is a more technical reason for this PR. Case conversion methods in String are among the largest functions in the standard library when measured by the number of lines of LLVM IR; #underscore measures 4.2k lines, #camelcase 1.9k lines, and #titleize 1.4k lines. This is due to inlining of the multiple { |c| io << c } blocks, which go all the way down to the body of e.g. Unicode.downcase(char : Char, options : CaseOptions, &). By establishing an inlining boundary, those line counts are now 0.67k, 0.23k, and 0.25k, whereas the new Char methods themselves contribute another 0.75k lines total.