Format Python code blocks in Markdown so ruff format --check passes (#112)

ruff 0.16.0 began formatting Python code blocks inside Markdown files. The
`dev` extra declares `ruff>=0.12.7` with no upper bound, so CI installs a ruff
that applies this, and the `check formatting` step now fails on two files that
have not otherwise changed. Because that step is first, it also gates linting,
type checking and the test suite, so builds stop before any of them run.

Reformats the two affected blocks only: a quote style in README.md and a
wrapped type annotation in circuit_tracer/utils/MAPPING_INFO.md. No prose,
no code semantics, and no public surface changes.

Author: speediedan

README.md

@@ -68,7 +68,7 @@ In order to use the `lazy_decoder` and `lazy_encoder` options on transcoders, th
 from circuit_tracer.utils.caching import save_transcoders_to_cache
 
 hf_ref = "mwhanna/gemma-scope-2-27b-pt/transcoder_all/width_262k_l0_small"
-cache_dir = '~/.cache/'
+cache_dir = "~/.cache/"
 save_transcoders_to_cache(hf_ref, cache_dir=cache_dir)
 ```
 

circuit_tracer/utils/MAPPING_INFO.md

@@ -12,7 +12,9 @@ class TransformerLens_NNSight_Mapping:
     embed_location: str  # Location of the embedding Module (the location to which we will attribute for embeddings)
     embed_weight: str  # Location of the embedding weight matrix
     unembed_weight: str  # Location of the unembedding weight matrix
-    feature_hook_mapping: dict[str, tuple[str, Literal['input', 'output']]]  # Mapping from (TransformerLens Hook) to a tuple representing an NNSight Envoy location, and whether we want its input or output
+    feature_hook_mapping: dict[
+        str, tuple[str, Literal["input", "output"]]
+    ]  # Mapping from (TransformerLens Hook) to a tuple representing an NNSight Envoy location, and whether we want its input or output
 ```
 
 The values of the location / pattern fields are the NNSight locations whose outputs we need to access or freeze. For example, if `embed_location="model.embed_tokens"`, this means that `model.model.embed_tokens` is the location of the embeddings, and `model.model.embed_tokens.output` will fetch its output.