Cxflow

Major features:
- `MainLoop` can be configured to `buffer` batches or produce ``fixed_epoch_size`` epochs (implemented with `StreamWrapper`)

Notable changes:
- `MainLoop` checks for stream methods before the training
- log dir contains docker-like sensible names
- **hooks**
- Add `LogDir` hook continuously logging the log dir path
- Add ETA and 1st epoch progress for `ShowProgress` hook

Notable changes:

- Add `DownloadableDataset` implementing common download and extract routines
- Add `ShowProgress` hook
- Rework tutorial and advanced sections in docs
- Minor fixes and test improvements

First **cxflow** release intended for public audience.

See https://cxflow.org for more information.

Major features
- Dataset, Model and Hook concepts
- MainLoop implementation
- **cxflow** command line tool with `train`, `resume`, `dataset` and `predict` commands
- **cxflow-tensorflow** adapter for TensorFlow https://tensorflow.cxflow.org
- **cxflow** examples available GitHub https://github.com/Cognexa/cxflow-examples
- **cxflow** hooks (https://cxflow.org/cxflow/cxflow.hooks)
- `WriteCSV`
- `StopAfter`
- `LogVariables`
- `LogProfile`
- `SaveEvery`
- `SaveBest`
- `CatchSigint`
- `ComputeStats`
- `Check`