Fix UTF-8 decoder rejecting valid U+0800 three-byte sequence

[metsw24-max]

This PR fixes an off-by-one error in the UTF-8 three-byte decoding validation logic.

Problem

Transcoder::decode incorrectly rejected the canonical UTF-8 encoding of U+0800.

The three-byte overlong validation check used:

if (rv <= 0x800)

which treated 0x0800 itself as invalid, even though it is the smallest valid code point that legitimately requires a three-byte UTF-8 sequence.

As a result, valid UTF-8 input containing the bytes:

E0 A0 80

(the canonical encoding of U+0800) was decoded as 0xFFFF, causing the caller to substitute Transcoder::LOSSCHAR and silently corrupt the decoded output.

Fix

Change the validation condition to:

if (rv < 0x800)

This preserves rejection of true overlong encodings while correctly accepting U+0800.

Tests

Added testDecodeUTF8_U0800 regression coverage which:

• decodes the literal UTF-8 byte sequence E0 A0 80
• verifies the decoded output matches Transcoder::encode(0x0800, …)
• asserts that no LOSSCHAR is introduced