Ruby: make sure map parsing frames are GC-rooted.

For JSON encoding we provide a new option to decide at
encode time whether to use camelCase or original proto field
names:

  json = MapMessage.encode_json(m, :preserve_proto_fieldnames => true)

This involved fixing a few important bugs in the
Ruby implementation -- mostly cases of mixing
upb field types and descriptor types (upb field
types do not distinguish between int/sint/fixed/sfixed
like descriptor types do).

Also added protobuf-specific exceptions so parse
errors can be caught specifically.

Change-Id: Ib49d3db976900b2c6f3455c8b88af52cfb86e036

Also fixed lines with > 80 char length.

- Alter encode/decode paths to use the `upb_env` (environment)
  abstraction.
- Update upb amalgamation to upstream `93791bfe`.
- Fix a compilation warning (void*->char* cast).
- Modify build flags so that upb doesn't produce warnings -- the Travis
  build logs were pretty cluttered previously.

* make consistent between mri and jruby
* create a #to_h and have it use symbols for keys
* add #to_json and #to_proto helpers on the Google::Protobuf message classes

This adds the Map container and support for parsing and serializing maps
in the protobuf wire format (as defined by the C++ implementation, with
MapEntry submessages in a repeated field). JSON map
serialization/parsing are not yet supported as these will require some
changes to upb as well.

system. The Ruby module build now uses an amalgamated distribution of
upb, and successfully builds a Ruby gem called 'google-protobuf' with
module 'google/protobuf'.

This adds a Ruby extension in ruby/ that is based on the 'upb' library
(now included as a submodule), and adds support for Ruby code generation
to the protoc compiler.