Fix four error messages that printed their placeholders (pytorch#23104)

### Summary

Four raised errors were plain strings, not f-strings, so users see the
literal placeholder instead of the value:

| File | Message |
|---|---|
| `exir/passes/quantize_io_pass.py` | `Output {output_index} is not a
dequantize op` |
| `exir/dialects/edge/dtype/runner.py` | `Type {t.name} does not have
dtype` |
| `backends/cortex_m/quantizer_reporter.py` | `Node {node.name} reported
as annotated but has no quantization annotation.` |
| `backends/arm/_passes/insert_table_ops.py` | `Table op for {target}
does not exist` |

For example, asking `QuantizeOutputs` to quantize an output that isn't a
dequantize op raises `ValueError: Output {output_index} is not a
dequantize op`. That doesn't say which output. Each fix adds the missing
`f` prefix, and no other code changes.

### Test plan

- Each changed line only adds an `f` prefix, and every variable it names
is in scope at the `raise`. All four files compile.
- `black` 24.4.2 and `flake8` 6.1.0, the lintrunner pins, pass on the
changed files. `ruff` 0.15.4 reports the same pre-existing findings as
`main`, all docstring rules, and nothing new.

This PR was authored with AI assistance (Claude Code). I found these
with a scan for string literals that contain `{name}` placeholders but
are never formatted.

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

cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils
@Sebastian-Larsson @robell @rascani

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

Author: David-Wu1119

backends/arm/_passes/insert_table_ops.py

@@ -130,7 +130,7 @@ def __getitem__(self, node: Node):
                     # Op must be handled if it's inside self.special_ops
                     raise AssertionError("Unhandled table operation")
         else:
-            raise KeyError("Table op for {target} does not exist")
+            raise KeyError(f"Table op for {target} does not exist")
 
     @staticmethod
     def included_ops() -> Iterator[EdgeOpOverload]:

backends/cortex_m/quantizer_reporter.py

@@ -130,7 +130,7 @@ def report_accept(self, pattern: List[Node]) -> None:
         for node in pattern:
             if Q_ANNOTATION_KEY not in node.meta:
                 raise ValueError(
-                    "Node {node.name} reported as annotated but has no quantization annotation."
+                    f"Node {node.name} reported as annotated but has no quantization annotation."
                 )
             annotation = cast(
                 QuantizationAnnotation | None, node.meta.get(Q_ANNOTATION_KEY)

exir/dialects/edge/dtype/runner.py

@@ -63,7 +63,7 @@ def mapping(t):
             elif t.is_scalar_type() or t.is_tensor() or t.is_tensor_list():
                 return type_dtypes + self.tensor_dtypes
             else:
-                raise ValueError("Type {t.name} does not have dtype")
+                raise ValueError(f"Type {t.name} does not have dtype")
 
         return list(map(mapping, types))
 

exir/passes/quantize_io_pass.py

@@ -165,7 +165,7 @@ def quantize_output(exported_program, output_index):
         exir_ops.edge.quantized_decomposed.dequantize_per_tensor.default,
         torch.ops.quantized_decomposed.dequantize_per_tensor.default,
     ]:
-        raise ValueError("Output {output_index} is not a dequantize op")
+        raise ValueError(f"Output {output_index} is not a dequantize op")
 
     dequant = target_output
     output_list[output_index] = dequant.args[0]