Datasiphon

- Validation changes: now instead of checking whether value in filter is of instance `column.type.python_type`, it uses the given `python_type` constructor to check if the value is processable by the given type
- Operations now apply the `python_type` constructor to the value before applying the operation

- Added advanced tests and logging

- Removed support for `ignore_extra_fields` and `strict` options in `sql.SQL.build` method
- added feature for infinite nesting - using `junctions` - AND/OR operators for sql queries - allowing advanced filtering and grouped clauses
- refactored validation of filter structure and columns, and restriction models
- added base class which has to be used when creating custom restriction models

- added option `strict` kwarg (defaults to `True`) to `sql.SQL.build` method to ignore fields in filter that don't correspond to expected `FORMAT` - does not apply to `ignore_extra_fields` option
- method `sql.SQL.validate_filter_structure` now instead of 'only validating' returns `dict` of processed filter items (because of `strict` option)

- added option `ignore_extra_fields` kwarg(defaults to `False`) to `sql.SQL.build` method to ignore extra fields in `filters` argument when using restriction model
- method `sql.SQL.validate_filter_columns` now instead of 'only validating' returns `dict` of processed filter items (because of `ignore_extra_fields` option)

- changed the way how `SQL` backend handles `order_by` keyword argument: now accepts `str` and `list` of `str` (instead of `dict` and `list` of `dict`) in following format options:
- `<order>(<field>)` - order being `asc` or `desc`
- `<order><field>` - order being `+` for `asc` and `-` for `desc`
- `<field>.<order>` - order being `asc` or `desc``