Fix JSON string-array serialization in TgTypeParser (closes #346)

[Dramex]

Closes #346.

The bug

Seven call sites passed parseArray<std::string> an identity lambda:

appendToJson(result, "emoji_list", parseArray<std::string>(
    [] (const std::string& s) -> std::string { return s; },  // <- identity
    object->emojiList));

The identity lambda produces [smile,heart] (unquoted identifiers) instead of ["smile","heart"]. That's not valid JSON, and @BloodRedTape confirmed in the issue thread that sticker creation is broken because of it. The same pattern exists in Api::setWebhook for allowed_updates.

The fix

Added a private helper in TgTypeParser:

void TgTypeParser::appendStringArrayToJson(std::string& json,
                                           const std::string& varName,
                                           const std::vector<std::string>& values) const;

which emits "name":["escaped","values"], directly, running each element through StringTools::escapeJsonString. It's a no-op when the vector is empty, matching the prior appendToJson(... parseArray(...)) behaviour.

Replaced the seven broken call sites in TgTypeParser.cpp:

• WebhookInfo.allowed_updates
• Chat.active_usernames
• Giveaway.country_codes
• InputSticker.emoji_list (the issue's primary reproducer)
• InputSticker.keywords
• PassportElementErrorFiles.file_hashes
• PassportElementErrorTranslationFiles.file_hashes

And fixed the single analogous site in Api::setWebhook by aligning its lambda with the pattern the other Api.cpp string-array sites already use (return "\"" + StringTools::escapeJsonString(s) + "\"").

What I did NOT change

I initially experimented with widening appendToJson(std::string)'s bracket heuristic to recognise [...] as pre-serialized JSON — this would have also fixed latent bugs in other array-wrapping call sites (e.g. parseMessage's entities). I reverted that because it would regress any field whose value legitimately starts with [ and ends with ] (keyboard-button text like "[CLICK]", captions, etc.). A dedicated helper is safer.

Only paths actually used by real Bot API requests are affected by this fix (setWebhook, createNewStickerSet/addStickerToSet via parseInputSticker, answerInlineQuery via the passport-error serializers).

Testing

Added test/tgbot/TgTypeParser.cpp with three regression tests:

• parseInputStickerEmojiListIsJsonArray — asserts the output contains "emoji_list":["smile","heart"] and not the pre-fix [smile,heart].
• parseInputStickerEscapesJsonSpecialsInEmoji — asserts inner " and \ get properly JSON-escaped.
• parseInputStickerOmitsEmptyKeywords — preserves prior behaviour where empty arrays are omitted from the output entirely.

Local build (Boost 1.90, OpenSSL 3.6, clang 17 on macOS arm64):

$ ./test/TgBot_test
Running 18 test cases...
*** No errors detected

(15 pre-existing + 3 new, all passing.)

Also verified end-to-end with a small smoke test:

Before: {"sticker":"file_id","format":"static","emoji_list":"[smile,heart]"}
After:  {"sticker":"file_id","format":"static","emoji_list":["smile","heart"]}