Unify decoding PE headers and reading other PE sections

[tmat]

PEHeaders are currently decoded using (customized) BinaryReader while the rest of the PE file is read using blob readers. We should use blob readers everywhere to avoid doing things in two different ways. As a by-product reading PE headers from non-zero position in the stream would just work. Currently it's broken.

[tmat]

See skipped test PEReaderTests.SubStream.

[karelz]

Be warned: This is rather tricky. You have to be careful about allocation lifetime, memory mapping, etc.

[danmoseley]

@tmat I bulk edited disabled tests to 2.0. Does this one actually need fixing in 2.0? It sounds like an enhancement. Can you change milestone if appropraite?