Broken text length and entity positions for a message that contain text with emojis

[unintended]

Context

• Operating System:
• Python Version: 3.7.7
• aiogram version: 2.9.2

if you send a message "🚀Hello Hello" to the bot and call message.parse_entities(False), you will get 🚀H*ello Hello*

Expected Behavior

message.parse_entities(False) returns 🚀*Hello Hello*

Current Behavior

message.parse_entities(False) returns 🚀H*ello Hello*

Failure Information (for bugs)

The problem is in json conversion and decoding of emoji surrogates. I debugged the protocol and getUpdates method returns the following json:
{"ok":true,"result":[{"update_id":337194970,"message":{"message_id":12146,"from":{"id":535059,"is_bot":false,"first_name":"\u0418\u043b\u044c\u044f","username":"igonzo","language_code":"en"},"chat":{"id":535059,"first_name":"\u0418\u043b\u044c\u044f","username":"igonzo","type":"private"},"date":1599124626,"text":"\ud83d\ude80Hello Hello","entities":[{"offset":2,"length":11,"type":"bold"}]}}]}
So the "text" has length 13 chars. And the bold entity starts at second char (offset=2) because emoji is encoded by two symbols (\ud83d\ude80). However after all the transformations the message object contains text of 12 unicode chars (because \ud83d\ude80 are joined into one 🚀char), but the entities are still 13-length based. After you try to apply the entities to the text (e.g. as it's done in message.parse_entities()), you get a wrong positioning.

Steps to Reproduce

Please provide detailed steps for reproducing the issue.

1. Run the bot with the following code:

import os

from aiogram import Bot, Dispatcher, executor, types
from aiogram.types import ContentType

bot = Bot(token=os.environ['BOT_TOKEN'])
dp = Dispatcher(bot)


@dp.message_handler(content_types=ContentType.ANY)
async def demarkdown(message: types.Message):
  await message.reply(message.parse_entities(False),
                      disable_web_page_preview=True)


if __name__ == '__main__':
  executor.start_polling(dp)

2. Send "🚀Hello Hello" to the bot (formatted)
3. You get 🚀H*ello Hello* replied

[unintended]

I tried to make a simple test using json.loads following by unwrapping the object to message and it works well.
Looks like there is some extra tuning in the bot, that joins surrogate unicode chars into emojis:

def test_parse_entities():
  upd_body = '{"ok":true,"result":[{"update_id":337194970,"message":{"message_id":12146,"from":{"id":535059,"is_bot":false,"first_name":"\u0418\u043b\u044c\u044f","username":"igonzo","language_code":"en"},"chat":{"id":535059,"first_name":"\u0418\u043b\u044c\u044f","username":"igonzo","type":"private"},"date":1599124626,"text":"\ud83d\ude80Hello Hello","entities":[{"offset":2,"length":11,"type":"bold"}]}}]}'
  upd_obj = json.loads(upd_body)
  assert len(upd_obj['result'][0]['message']['text']) == 13

  msg = Message(**upd_obj['result'][0]['message'])
  assert len(msg.text) == 13
  assert msg.parse_entities(False) == "\ud83d\ude80*Hello Hello*"

[unintended]

Also this problem doesn't occur in MessageEntity.get_text() because the text is converted to 'utf-16-le' before applying the offsets, maybe it's a good point to start with.

[unintended]

I made a pretty raw fix #414 that uses the same logic as it's done in MessageEntity.get_text() to resolve this issue. May need a rework though.