Fix Sophia neural network training documentation example

Addresses issue #937 by fixing the loss function in the Sophia training example to properly handle DataLoader batches.

The issue was in the loss function which was trying to access `data[1]` and `data[2]` directly on a DataLoader object, which is not indexable in that way. The corrected version:

- Changes parameter name from `data` to `batch` for clarity
- Properly unpacks the batch using `x_batch, y_batch = batch`
- Uses the unpacked batch data correctly in the loss computation

This follows the same pattern as the working minibatch tutorial and allows users to run the example outside of the optimization loop, as requested in the issue.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>

Author: ChrisRackauckas

docs/src/optimization_packages/optimization.md

@@ -80,9 +80,10 @@ function callback(state, l)
     return l < 1e-1 ## Terminate if loss is small
 end
 
-function loss(ps, data)
-    ypred = [smodel([data[1][i]], ps)[1] for i in eachindex(data[1])]
-    return sum(abs2, ypred .- data[2])
+function loss(ps, batch)
+    x_batch, y_batch = batch
+    ypred = [smodel([x_batch[i]], ps)[1] for i in eachindex(x_batch)]
+    return sum(abs2, ypred .- y_batch)
 end
 
 optf = OptimizationFunction(loss, AutoZygote())