OpenJDK HPROF dump parsing library.

Based on https://hg.openjdk.java.net/jdk/jdk/file/9a73a4e4011f/src/hotspot/share/services/heapDumper.cpp
and http://hg.openjdk.java.net/jdk6/jdk6/jdk/raw-file/tip/src/share/demo/jvmti/hprof/manual.html

Currently, the supported ID sizes (actually, pointer sizes) are 4 and 8.

The library provides a streaming interface (iterator over dump
records).  Class structure information and string dictionary are kept
in memory as they are required for unpacking object data.

There are two modes of operation: reading from std::io::BufRead with
StreamHprofReader::read_hprof_from_stream (strings are
returned as Vec<u8>) and from memory with
StreamHprofReader::read_hprof_from_memory (strings are returned as
&'memory [u8], where 'memory is memory lifetime).

Values stored in Record::String (i.e. strings by themselves) do not always
have valid UTF-8 data, that's why they are returned as byte vectors/slices.

Strings are returned as Record::String(id, data), and
from object dump, these strings are refered by their id; this
crate doesn't resolve these ids into strings (and objects are not
resolved either), you have to keep id->data mapping by yourself.

Author: Ivan Boldyrev <lispnik@gmail.com>