BigQuery loading via GCS

Breaking changes:

• None

New features:

• BigQuery import from GCS buckets (#113)
• Allow Parquet records format to be specified on mvrec command line (#129)
• Allow environment-based configuration of GCS creds (#120)
• Do slow redshift unload via SELECT when bucket unload not available (#117)
• Load via INSERT on Redshift when scratch bucket not available (#114)

Bug fixes / reliability improvements:

• Prefer file extension to dialect compression defaults in targets (#111)
• Handle multiple fileobjs in DoMoveFromFileobjsSource (#107)
• Fix README.md code sample errors (#106)
• Also downcast constraints and statistics when downcasting field types (#103)
• Add dependency fix for sudden BigQuery test failure (#109)
• Allow Parquet to be used in import to BigQuery from records directories (#130)
• Handle 'operation timed out error' during long Redshift unloads (#128)
• Retry on more Google rate limit exceptions (#126)
• Better error message when S3 _format_* file doesn't exist. (#121)
• Improve logging for large moves (#122)
• Update PyYAML dependency to match awscli (#102)

Other updates:

• Fix dependencies for Homebrew processing (#135)
• Rename some internally used methods (#124) (#116)
• Drop dead code (#123) (#115)
• Rename module for better Mypy support (#125)
• Also test Redshift without S3 scratch bucket (#118)
• Introduce component test suite (#108)
• Bump Python version for internal development (#100)