Fix ESC/POS notation parser: '#' only starts a comment as its own token

Stripping from the first '#' anywhere in the line ate content like '#5
Fix thing' (an issue reference) entirely, since it starts with '#' with
no preceding whitespace. Comments now only trigger on a standalone '#'
token (i.e. whitespace on both sides), matching the documented examples.

app.rb

@@ -117,17 +117,23 @@ helpers do
   #   Hello          # literal text
   #   LF
   #   GS k 0x03 12345678 0x00
-  # Each `#...` trailing comment is stripped, then each line's
+  # A trailing comment is only a comment when "#" is its own
-  # whitespace-separated tokens become either a known control mnemonic, a
+  # whitespace-separated token -- not just stripped from the first "#"
-  # raw `0xNN` byte, or literal text (CP852-encoded) -- consecutive literal
+  # anywhere in the line, which would mangle literal text like "#5 Fix
-  # tokens are rejoined with a single space so plain prose still reads
+  # thing" (an issue reference, not a comment). Each remaining token
-  # correctly, while whitespace around control/byte tokens is dropped since
+  # becomes a known control mnemonic, a raw `0xNN` byte, or literal text
-  # those are just for human readability, not printer output.
+  # (CP852-encoded) -- consecutive literal tokens are rejoined with a
+  # single space so plain prose still reads correctly, while whitespace
+  # around control/byte tokens is dropped since those are just for human
+  # readability, not printer output.
   def parse_escpos_notation(text)
     bytes = "".b
     text.each_line do |line|
       prev_was_text = false
-      line.sub(/#.*/, '').split.each do |token|
+      tokens = line.split
+      comment_at = tokens.index('#')
+      tokens = tokens[0...comment_at] if comment_at
+      tokens.each do |token|
         if ESCPOS_MNEMONICS.key?(token)
           bytes << ESCPOS_MNEMONICS[token]
           prev_was_text = false