The implementations of Module/UnloadedModule and
ModuleList/UnloadedModuleList are very similar. They have been made
separate classes because they operate on different structs, complicating
factoring code into a base class and have sufficiently different
implementation that templates would not be suitable.

When unloaded modules have partially overlapping ranges, the module
shrink down feature is used to move the start of the higher range to the
end of the lower range. If two unloaded modules overlap identically, the
second module will not be added to the range map and the failure
ignored.

Places where MinidumpUnloadedModule differs from MinidumpModule:
  code_identifier: the android/linux case is deleted since cv_records
    never exist.
  debug_file/debug_identifier/version: always return empty strings.
  Read: an expected size is provided as opposed to MD_MODULE_SIZE. A
    seek is used if there are extra, unused bytes.

Places where MinidumpUnloadedModuleList differs from
  MinidumpModuleList:
  Read: entry and header size is provided in the header in
    addition to count. This changes the checks and handling of padding.
    Failures from StoreRange are ignored.
  GetMainModule: always returns NULL.

BUG=

Change-Id: I52e93d3ccc38483f50a6418fede8b506ec879aaa
Reviewed-on: https://chromium-review.googlesource.com/421566


Reviewed-by: Joshua Peraza <jperaza@chromium.org>