Description:
------------
When calling iconv_mime_decode() with $mode = ICONV_MIME_DECODE_CONTINUE_ON_ERROR, 
the manual states that "iconv_mime_decode_headers() attempts to ignore any 
grammatical errors and continue to process a given header."

Accordingly, B-encoded encoded-words (as per RFC 2047) are silently ignored.

But when using Q (quoted-printable) encoding, encoded-words with illegal 
characters (> 127) cause iconv_mime_decode() to return false as if 
ICONV_MIME_DECODE_CONTINUE_ON_ERROR was not specified. 

As such it is not resilient towards malformed headers. 

Test script:
---------------
$m = ICONV_MIME_DECODE_CONTINUE_ON_ERROR;

var_dump(iconv_mime_decode("Legal encoded-word: =?utf-8?B?Kg==?= .", $m));
var_dump(iconv_mime_decode("Legal encoded-word: =?utf-8?Q?*?= .", $m));
var_dump(iconv_mime_decode("Illegal encoded-word: =?utf-8?B?".chr(0xA1)."?= .", $m));
var_dump(iconv_mime_decode("Illegal encoded-word: =?utf-8?Q?".chr(0xA1)."?= .", $m));

Expected result:
----------------
string(23) "Legal encoded-word: * ."
string(23) "Legal encoded-word: * ."
string(24) "Illegal encoded-word:  ."
string(24) "Illegal encoded-word:  ."

Actual result:
--------------
string(23) "Legal encoded-word: * ."
string(23) "Legal encoded-word: * ."
string(24) "Illegal encoded-word:  ."
bool(false)