fix(interpreter): decode each statement's stdout via stdoutKind to fix UTF-8 mojibake (#272)

A single exec() can interleave text-shaped statements (sed, awk, echo — ö as
U+00F6) with byte-shaped ones (grep | head, cat — ö as bytes 0xC3 0xB6).
executeScript / executeStatement concatenated each result's raw stdout, so the
lone high byte from the text half made the combined stream invalid UTF-8, the
output-boundary decoder bailed, and the byte half came back as Latin-1 mojibake
(Köpenicker for Köpenicker). The same path backs command substitution, so
echo "…: $(cat /file)" was affected too.

Decode each statement/pipeline result to text via its explicit stdoutKind
(decodedTextFromResult) before concatenating — no guessing from string
contents, so text whose code units merely look like UTF-8 (ö) is preserved.
The shape is declared per output rather than per command: tac (stdin path) and
curl (response body) now set stdoutKind: "bytes" on the results that forward
raw bytes; the rest already self-mark via stdoutKind / the legacy
stdoutEncoding alias.

Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>