Actually, you can just place a % symbol every two characters. Azureus doesn't do that because, for example, R is a safe character in a URL, and 52 is the hexadecimal representation of R, so it doesn't need to percent-encode it. Using %52 instead is equivalent.

Even though I know well the original question was about C++, it might be useful somehow, sometimes to see alternative solutions. Therefore, for what it's worth (10 years later), here's

An alternative solution implemented in Python 3.6+

import binascii
import urllib.parse

def hex_str_to_esc_str(s: str, *, encoding: str='Windows-1252') -> str:
    # decode hex string as a Windows-1252 string
    win1252_str = binascii.unhexlify(hex_str).decode(encoding)
    # escape string and return
    return urllib.parse.quote(win1252_str, encoding=encoding)

def esc_str_to_hex_str(s: str, *, encoding: str='Windows-1252') -> str:
    # unescape the escaped string as a Windows-1252 string
    win1252_str = urllib.parse.unquote(esc_str, encoding='Windows-1252')
    # encode string, hexlify, and return
    return win1252_str.encode('Windows-1252').hex()

Two elementary tests:

esc_str = '%D9%0C%3C%E3%94%18%F0%C5%D9%83X%E03I%26%2B%60%8C%BFR'
hex_str = 'd90c3ce39418f0c5d98358e03349262b608cbf52'

print(hex_str_to_esc_str(hex_str) == esc_str) # True
print(esc_str_to_hex_str(esc_str) == hex_str) # True

Note

Windows-1252 (aka cp1252) emerged as the default encoding as a result of the following test:

import binascii
import chardet

esc_str = '%D9%0C%3C%E3%94%18%F0%C5%D9%83X%E03I%26%2B%60%8C%BFR'
hex_str = 'd90c3ce39418f0c5d98358e03349262b608cbf52'

print(
    chardet.detect(
        binascii.unhexlify(hex_str)
    )
)

...which gave a pretty strong clue:

{'encoding': 'Windows-1252', 'confidence': 0.73, 'language': ''}