Keep characters after a truncated UTF-8 sequence (#3690)

A multibyte UTF-8 sequence cut short by a non-continuation byte took
every remaining byte of the string, so "a\xf0Zb" became "a\ufffd".

Replace only the lead byte and the continuation bytes that follow it,
the same as an invalid sequence that is not cut short, so "a\xf0Zb"
now becomes "a\ufffdZb". This also changes length and explode on such
strings, which now count the characters that used to be dropped.

tests/jq.test

@@ -89,6 +89,11 @@ map(@base64d)
 ["YWJjZGVmZ2hpav9rbG1ub3A=", "YWJjZGVmZ2hpasOpa2xtbm9w5pel", "/2FiY2RlZmdoaWo=", "YWJjZGVmZ2hpauaX", "YWJjZGVmZ2hpau2ggHh5eg=="]
 ["abcdefghij�klmnop","abcdefghijéklmnop日","�abcdefghij","abcdefghij�","abcdefghij�xyz"]
 
+# Truncated UTF-8 sequence keeps the characters after it
+map(@base64d)
+["8Fo=", "4mE=", "8J9a", "YfBaYg==", "8J+Y"]
+["�Z","�a","�Z","a�Zb","�"]
+
 @uri
 "\u03bc"
 "%CE%BC"