Feature transform api

[PhilippvK]

See #81

Changes:

• Refactor transforms
  • Move argparse handling to get_parser functions
  • Rename main to run which expects argparse namespace
  • Introduce new main handling cmdline args
  • Implement "fake"-class for transforms in __init__.py. Example: from .transform import main as InferTypes
• Update Passes
  • Add use_subprocess argument to each transform pass
  • Use new "fake" class based API for transforms. Example:

    from seal5.transform.infer_types import InferTypes
    args = sanitize_args(args)
    InferTypes(args)

  • Expose transform log level to each pass settings
• Examples/Configs
  • Set use_subprocess and log_level for transform passes

TODOs:

• [ ] Apply these changes also to frontends and backends
• [ ] Allow adding global pass defaults (instead of configuring each pass separately)
• [x] Resolve lint issues

[PhilippvK]

@thomasgoodfellow Feel free to have a look at this. This is a minimal "hack" to expose the transforms to be called from python while still using the "cmdline-fist" approach. The migration to a more powerful api (which will especially be helpful for sharing common code) can be done later as proposed in #54