Revert "PrettyPrinterPerformance Optimized the PrettyPrinter for #894"

[bnbarham]

Reverts #901 which now computes the column on codepoints rather than characters.

[bnbarham]

CC @macshome (sorry, we'll get there!)

I haven't benchmarked this at all, but we could do something like:

    guard let lastNewlineIndex = text.lastIndex(of: "\n") else {
      // In case there's no newline, the string itself is the only line
      column += text.count
      return
    }

    let lastLine = text[lastNewlineIndex...]
    column += lastLine.count

    // Now count the rest of the lines in a possible multi-line string. We are only interested in '\n' so we can use
    // the UTF8 view and skip grapheme clustering entirely.
    for element in text[...lastNewlineIndex].utf8 {
      if element == UInt8(ascii: "\n") {
        lineNumber += 1
      }
    }

Which should hopefully result in a single iteration still (though I'm unsure of the slices). But basically - iterate backwards to find the last newline, grab column count during/from there, then iterate forwards up to that index to get the other newlines (which can still use the utf8 view).

Couple extra tests we could add:

  func testCharacterVsCodepoint() {
    let input =
      """
      let fo = 1  // 🤥
      
      """

    assertPrettyPrintEqual(
      input: input,
      expected: input,
      linelength: 16,
      whitespaceOnly: true,
      findings: []
    )
  }

  func testCharacterVsCodepointMultiline() {
    let input =
      #"""
      /// This is a multiline
      /// comment that is in 🤥
      /// fact perfectly sized
      
      """#

    assertPrettyPrintEqual(
      input: input,
      expected: input,
      linelength: 25,
      whitespaceOnly: true,
      findings: []
    )
  }

[ahoppen]

@macshome Would you be interested in trying something like what @bnbarham suggested to count characters instead of codepoints?