BUG: Use micro average for MultiClassAccuracy (#671)

the default for MultiClassAccuracy is macro [Accuracy — PyTorch-Metrics
0.10.3 documentation
(torchmetrics.readthedocs.io)](https://torchmetrics.readthedocs.io/en/stable/classification/accuracy.html#multiclassaccuracy)

But base class (which we were using before) is micro [Accuracy —
PyTorch-Metrics 0.10.3 documentation
(torchmetrics.readthedocs.io)](https://torchmetrics.readthedocs.io/en/stable/classification/accuracy.html)

We need to set average=micro

hi-ml-cpath/src/health_cpath/models/deepmil.py

@@ -176,7 +176,7 @@ class BaseDeepMILModule(LightningModule):
     def get_metrics(self) -> nn.ModuleDict:
         if self.n_classes > 1:
             return nn.ModuleDict({
-                MetricsKey.ACC: MulticlassAccuracy(num_classes=self.n_classes),
+                MetricsKey.ACC: MulticlassAccuracy(num_classes=self.n_classes, average='micro'),
                 MetricsKey.AUROC: MulticlassAUROC(num_classes=self.n_classes),
                 MetricsKey.AVERAGE_PRECISION: MulticlassAveragePrecision(num_classes=self.n_classes),
                 # Quadratic Weighted Kappa (QWK) used in PANDA challenge