echo escape conformance: xpg_echo + the \0nnn octal rule (batch83)

[brianjfox]

Summary

Two related echo escape-handling fixes surfaced by builtins2.sub (builtins byte-diff 179 → 172, no regressions).

1. echo octal is \0nnn, not a bare \nnn

bash's echo (with -e, and with xpg_echo) accepts only the \0nnn octal escape; a bare \1..\7 stays literal — unlike printf %b, which interprets \nnn. gnash's shared decode_b interpreted the bare form in both, so echo -e "\101" printed A instead of literal \101. Added a bare_octal parameter (default true, preserving printf %b) and pass false from echo.

2. honor the xpg_echo shopt

shopt -s xpg_echo now makes echo interpret backslash escapes by default (as if -e), matching bash (do_v9 = xpg_echo). In POSIX mode with xpg_echo, option parsing is skipped entirely so -n/-e/-E print literally (bash's posixly_correct && xpg_echo short-circuit). gnash previously recognized the shopt name but ignored it.

$ shopt -s xpg_echo; echo 'a\tb'      →  a<TAB>b
$ echo -e '\101 \0101'               →  \101 A
$ printf '%b\n' '\101'               →  A

Testing

• ctest 22/22
• Scoreboard: builtins 179 → 172, zero regressions across the full suite
• builtins2.sub now matches bash byte-for-byte; verified octal/hex/xpg/posix cases against bash 5.3

Closes #283.