Fix decoding UTF-8 constant pool entries

[Ladicek]

Constant pool entries of type CONSTANT_Utf8_info do not, despite
the name, use the UTF-8 encoding. They use a modified variant
of UTF-8, as specified by java.io.Data{Input,Output}.

When decoding these entries, the Indexer.decodeUtf8Entry method
interpreted the data as UTF-8. This didn't cause issues, because
the constants are usually human-readable strings that fit the ASCII
table, in which case the two encodings do not differ.

In case of machine-generated content, the difference may easily
occur; for example in case of a string that contains the null
character. One realistic example is the Kotlin standard library
JAR, where the kotlin/collections/ArraysKt___ArraysKt.class class
contains a @KotlinMetadata annotation whose d1 member contains
such "weird" string.

The fix is simple: use DataInputStream to read a String out of
the byte array.

Fixes #49