Fix #1112: ASCII85 decode now handles CRLF line endings

The ASCII85 decoder previously failed when encoded data had both
carriage return and newline (CRLF) at the end. The code only
stripped one trailing character, causing the EOD marker check to
fail when CRLF was present.

Changed to use bytes.TrimRight to strip all trailing CR/LF
characters, which aligns with the PDF specification's requirement
to ignore whitespace in ASCII85 encoding.

Added test case to verify decoding works with LF, CR, CRLF, and
no line endings.