reverted update where all Dataset output examples to use python synta…

…x highlighting, as this makes it harder to detect the difference between code and output in the docs, and is also less true to reality, as outputs are not syntax highlighted.

benchmarking/datasets.mdx

@@ -37,7 +37,7 @@ then this is interpreted as the user message
 import unify
 datum = unify.Datum("This is a user message.")
 ```
-```python
+```
 Datum(
     prompt=Prompt(
         messages=[{'content': 'This is a user message.', 'role': 'user'}],
@@ -66,7 +66,7 @@ Datum(
 Again, this is better visualized after setting `unify.set_repr_mode("concise")`.
 We will assume `"concise"` mode for the rest of the examples on this page:
 
-```python
+```
 Datum(
     prompt=Prompt(
         messages=[{'content': 'This is a user message.', 'role': 'user'}]
@@ -85,7 +85,7 @@ datum = unify.Datum(
     ref_answer="This is an answer."
 )
 ```
-```python
+```
 Datum(
     prompt=Prompt(
         messages=[{'content': 'This is a user message.', 'role': 'user'}]
@@ -112,7 +112,7 @@ When a list of strings is passed, these strings are also interpreted as *user me
 but as explained above, the underlying dataset will contains `Datum` instances,
 which themselves contain `Prompt` instances:
 
-```python
+```
 Dataset(
     [
         Datum(
@@ -153,7 +153,7 @@ dataset = unify.Dataset([
     ),
 ])
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -208,7 +208,7 @@ dataset = unify.Dataset([
     ),
 ])
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -243,7 +243,7 @@ interpreted as single-item datasets:
 import unify
 dataset = unify.Dataset("First user message.")
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -296,7 +296,7 @@ dataset = unify.Dataset([
 ])
 print(dataset[1:-1])
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -344,7 +344,7 @@ dataset2 = unify.Dataset([
 ])
 print(dataset1 + dataset2)
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -385,7 +385,7 @@ dataset2 = unify.Dataset([
 ])
 print(dataset1 + dataset2)
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -434,7 +434,7 @@ dataset += "Second user message."
 dataset += unify.Prompt(messages=[{'content': 'Third user message.', 'role': 'user'}])
 dataset += ["Fourth user message.", "Fifth user message."]
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -473,7 +473,7 @@ The ordering of the operation also does not matter, by virtue of `__radd__` and
 import unify
 "User message" + unify.Dataset("Another user message.")
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -502,7 +502,7 @@ dataset = "First user message." \
            + unify.Datum("Third user message")
 print(dataset)
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -542,7 +542,7 @@ dataset2 = unify.Dataset([
 ])
 print(dataset1 - dataset2)
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -592,7 +592,7 @@ dataset -= "First user message"
 dataset = dataset - unify.Prompt("Second user message.")
 print(dataset)
 ```
-```python
+```
 Dataset(
     [
         Datum(
@@ -697,7 +697,7 @@ You can directly download a dataset from your account as follows:
 import unify
 dataset = unify.Dataset.from_upstream("my_dataset")
 ```
-```python
+```
 Dataset(
     [
         Datum(