Data2neo

This update brings big performance improvements by adding two main improvements:

- Resources that do not produce a node or a relation are bypassed which resolves a large bottleneck in the old code
- Intermediate subgraphs that are produces for each resource are now newly buffered. The size of the buffer is dynamically updated based on the throughput of resources.

These improvements bring significant performance benefits (1.5-3 times faster processing)

Further the `py2neo_extensions` now provide a wrapper that enables matching of relations with no primary key between the same two nodes.

The converter newly only accepts schemas as a string. This allows the schema string to be generically generated. Futher the library provides a `load_file` function under `rel2graph.utils` to make switching to the new api as easy as possible.

Old way:
python
converter = Converter(path_to_file, my_iterator, my_graph)

New way:
python
from rel2graph.utils import load_file
converter = Converter(load_file(path_to_file), my_iterator, my_graph)

This release fixes a bug where primary keys were not correctly parsed if the primary attribute was wrapped.

This release fixes all known issues with the schema compiler, further the compiler gives know much more meaningful error messages.