fix(write): named GPIO write values (on/high/true) were silently driving pins LOW (gap-hunt round 2)

[SuperInstance]

Gap-hunt pass (goose), independently re-verified by me (rebuilt firmware + test env on the exact pushed commit, hand-traced the fixed logic against the header's documented interface — no real ESP32 hardware is available in this environment, so like round 3's original verification for this repo, this is compile-clean + logic-traced, not device-executed).

Bug: executeWrite() used valueStr.toInt() to decide the GPIO level. On Arduino, String::toInt() returns 0 for any non-numeric input — so write <pin> on, write <pin> high, and write <pin> true all silently drove the pin LOW, directly contradicting the documented interface (both the README and the help text in CommandParser.cpp advertise on/off, high/low as valid values). A real, functional bug: any deployment following the repo's own documented usage for named values would get the opposite of the requested pin state.

Fix: explicitly parses the named aliases (on/high/true/1 -> HIGH, off/low/false/0 -> LOW, case-insensitive) before falling back to numeric parsing — matching the contract already implemented (but previously unused for this path) by parseValue() in CommandParser.cpp.

Verification (independently redone):

• Pulled the exact pushed commit, ran pio run -e esp32dev (SUCCESS) and pio test -e test --without-uploading --without-testing (compiles clean) myself.
• Hand-traced the new boolean logic against all 8 named aliases plus 0/1: every truthy alias (on,high,true,1) resolves to HIGH, every falsy alias (off,low,false,0) resolves to LOW — matches the new regression test (test_write_named_values) exactly.
• Note (same caveat as round 3's original verification of this repo): no real ESP32 device is available in this environment, so Unity test assertions aren't actually executed here, only compiled — this is the same verification standard already established and accepted for this repo.