Add additional benchmark and model metadata

[mike-ferguson]

This PR adds metadata fields to both models and benchmarks with known reference paper URL. The fields task_specialization and training_dataset were added to 28 models (with papers), while benchmarks add in task, species, region, and datatype.

More specifically for models:

Models with training_dataset: 28
Models with task_specialization: 28

======================================================================
TRAINING DATASETS
======================================================================
  training_dataset: ImageNet-1k: 21 models
  training_dataset: Tiny_ImageNet: 6 models
  training_dataset: CIFAR-100: 2 models
  training_dataset: ImageNet-22k: 2 models
  training_dataset: VTAB: 1 model
  training_dataset: custom_dataset: 1 model

======================================================================
TASK SPECIALIZATIONS
======================================================================
  task_specialization: image_classification: 9 models
  task_specialization: brain_alignment: 8 models
  task_specialization: scaling_models: 7 models
  task_specialization: object_detection: 5 models
  task_specialization: class_activation_mapping: 1 model
  task_specialization: graph_recognition: 1 model
  task_specialization: mobile_vision: 1 model
  task_specialization: salient_object_detection: 1 model

And benchmarks:

Total benchmarks found: 184
Benchmarks with benchmark_type: 184
Benchmarks with task: 76
Benchmarks with region: 56
Benchmarks with species: 137
Benchmarks with datatype: 133

======================================================================
BENCHMARK TYPES
======================================================================
  benchmark_type: behavioral: 84 benchmarks
  benchmark_type: neural: 56 benchmarks
  benchmark_type: engineering: 44 benchmarks

======================================================================
TASKS
======================================================================
  task: alternative_forced_choice: 39 benchmarks
  task: passive_viewing: 37 benchmarks

======================================================================
REGIONS
======================================================================
  region: V1: 29 benchmarks
  region: IT: 16 benchmarks
  region: V4: 12 benchmarks
  region: V2: 5 benchmarks

======================================================================
SPECIES
======================================================================
  species: homo_sapiens: 113 benchmarks
  species: macaca_mulatta: 32 benchmarks

======================================================================
DATATYPES
======================================================================
  datatype: behavioral: 84 benchmarks
  datatype: engineering: 44 benchmarks
  datatype: fMRI: 4 benchmarks
  datatype: neural: 1 benchmark

Note: this is a first step in adding metadata. Benchmark metadata requires a review by a human before merge. Model metadata has largely been confirmed by me.

Note: I am did not conduct this work, I have only validated models; all credit to Nishka Shah