Add a lookahead buffer to TokenList::Stream #739

TokenList::Stream turns the input bytes into characters: it decodes UTF-16, and it turns CR and CR LF into LF. Putting a character
back (ungetChar()) is passed down to the stream classes as raw byte unget() calls:

So only one character can safely be put back, and ungetChar() must undo the UTF-16 decoding byte by byte (C only guarantees one
character of ungetc() pushback).

Step 2 of #708 moves line splicing into the stream. Then readChar() must look past \, any spaces, and an optional CR to see if a
newline follows. If it doesn't, all of those characters must be put back.

Add a small buffer of already decoded characters to Stream:

• readChar() and peekChar() take characters from the buffer first.
• ungetChar() puts the last character back into the buffer, instead of calling unget() on the stream classes.
• More than one character can be put back.

files with CR-only line endings are misread with file input, and a CR at the end of the file is not treated as a newline.