PumpReader.createInputStream(...) returns EOF for supplementary code points #658

The InputStream created using org.jline.utils.PumpReader.createInputStream(Charset) returns EOF (-1) when encountering a supplementary code point (i.e. > U+FFFF) in the input, e.g.:

PumpReader reader = new PumpReader();
reader.getWriter().append("\uD83D\uDE0Atest");
InputStream in = reader.createInputStream(StandardCharsets.UTF_8);
in.read();

The reason for this is that the buffer is sized incorrectly here:

It appears often encoders try to encode supplementary code points (represented by a surrogate pair consisting of two char) either completely or not at all, so the buffer with size encoder.maxBytesPerChar() is too small, using 2 * encoder.maxBytesPerChar() should solve this issue.

However, even when this is fixed, it would be good to improve the encoding logic by checking for OVERFLOW (instead of silently ignoring it) and throwing an AssertionError here: