updates for v1.6

[jpata]

• Improve the training and evaluation recipe for the benchmark model provided in the paper. Addresses readme for using the pretrained model #217
• fix data preprocessing fails with latest awkward #222
• fix CUDA optimization prevents running on older cards #220
• if padded batching is used, compute number of steps naively, do not step through the dataset to get the number of steps
• split the Delphes dataset clearly to ttbar and qcd, like the other datasets
• regenerate the Delphes dataset using tfds with array_record
• fix re-enable ONNX export in tensorflow #225

Validation:

• delphes training
• clic cluster-based training
• cms training

Re-running the inference on the pretrained CLIC cluster-based weight file https://huggingface.co/jpata/particleflow/blob/main/weights-96-5.346523.hdf5 gave the expected result (this plot was made with limited stats):