Debugger: spawn monsters by id, and fix truncated UTF-8 names

[devtejasx]

Two small debugger changes, one commit each.

Spawn by id (#113). The unit spawner only had a combo box of names, so reaching a given monstats or superuniques row meant scrolling through hundreds of entries, and rows that share a name (lots of "Fallen", "Zombie", ...) couldn't be told apart.

• The combo boxes now show <id>: <name>.
• Each spawner gets an id input that selects the row directly, clamped to the table's record count.

This covers the "SuperUnique monster by id" and "Normal monster by id" items in #113.

Truncated names. GetUTF8CharBufferFromStringIndex sized its buffer at exactly 3 bytes per character and passed size - 2 to toUtf, which stops once count - 1 bytes are used. So:

• a 1-character name came back empty,
• names of 3-byte characters lost their last character (中文 became 中),
• an empty string made an empty vector, so the count was size_t(-2) and toUtf wrote its terminator through the null data() pointer.

The buffer now has 3 extra bytes, enough for the last character and never empty.

I can't build the project here (no 32-bit MSVC), so I checked both parts outside it: the new helpers compile with g++ -Wall -Wextra against a stub ImGui::InputInt and give the expected labels and clamping. A host model of toUtf's loop converts "", "A", "Ab", "Andariel", "é", "中" and "中文" in full with the new size, and reproduces all three problems with the old one.

Refs #113