where "【" should be 12304 when encoded but it become splited into 3 parts: 227,128,144

Yes, that is properly encoded UTF-8. Codepoints from U+0800 to U+FFFF are to be encoded with 3 bytes. The codepoint 12304, which is 0x3010 in hex, usually using the U+3010 notation for Unicode, should be encoded as the three bytes 0xE3, 0x80, 0x90. Working it out:

Codepoint 12304 (codepoint 0x3010 in hex, U+3010)

hex 0x3010
hex 3    0    1    0
bin 0011 0000 0001 0000
    xxxx yyyy yyzz zzzz     (use x, y, and z to indicate the groups of bits in the codepoints)

encoding:
    ....xxxx ..yyyyyy ..zzzzzz  (use xyz as above; use dots . to indicate bits specified in UTF-8 encoding)
bin 11100011 10000000 10010000

hex E3       80       90
dec 227      128      144

... which is what you listed

(This is, btw, why Corion told you to look for the charset=utf-8 in the Content-type, because he recognized those three bytes were the appropriate UTF-8 encoding of the LEFT BLACK LENTICULAR BRACKET (U+3010) )