docs(design): address CodeRabbit review (typo + code-fence language tags)

CodeRabbit round 1 on PR #637:

- s3_raft_blob_offload.md:358 typo "elegibility" → "eligibility"
  (single character).
- Add `text` language identifier to the six previously-bare fenced
  code blocks (the §1 worst-case formula in admission_control,
  the §3.3 client/admission flow diagram in admission_control,
  the §3.5 metrics block in admission_control, the §1
  Raft-log flow in blob_offload, the §3.1 keyspace block in
  blob_offload, and the §3.2 PUT pipeline diagram in
  blob_offload). Markdownlint flags bare fences; tagging them
  enables both linting compliance and consistent syntax
  highlighting in the rendered docs.

No code changes; design docs only.

Author: bootjp

docs/design/2026_04_25_proposed_s3_admission_control.md

@@ -22,7 +22,7 @@ with `MaxSizePerMsg`, the **aggregate** in-flight memory bound is
 governed by client concurrency, which the server cannot currently
 limit:
 
-```
+```text
 leader-side worst-case  =  concurrent_PUTs × pending_entries_per_PUT × entry_size
                         ≈  concurrent_PUTs × MaxInflight × 4 MiB
 ```
@@ -105,7 +105,7 @@ Recommended: **both, at different scales**.
   fails with 503 mid-stream. Dispatch latency stays bounded by the
   semaphore size rather than by client behaviour.
 
-```
+```text
 client ─[Content-Length]─► (A) reserve full body bytes
                                 │
                                 ▼
@@ -323,7 +323,7 @@ materialise by default.
 
 ### 3.5 Metrics
 
-```
+```text
 elastickv_s3_put_admission_inflight_bytes        gauge
 elastickv_s3_put_admission_rejections_total      counter (labels:
                                                     stage    = "prereserve" | "perbatch",

docs/design/2026_04_25_proposed_s3_raft_blob_offload.md

@@ -20,7 +20,7 @@
 
 Today every byte of an S3 object travels through the Raft log:
 
-```
+```text
 HTTP PUT body  ─►  s3ChunkSize (1 MiB) chunks
                  ─►  s3ChunkBatchOps × 1 MiB Raft entry
                  ─►  Raft log entry (Pebble WAL on every node)
@@ -102,7 +102,7 @@ themselves*.
 
 ### 3.1 New keyspace
 
-```
+```text
 !s3|chunkref|<bucket>|<gen>|<objectKey>|<uploadID>|<partNo>|<chunkNo>
     → ChunkRef{
         ContentSHA256 [32]byte
@@ -128,7 +128,7 @@ Two separate keyspaces:
 
 ### 3.2 PUT path
 
-```
+```text
 client ─► HTTP PUT body
         ─► chunk loop (s3ChunkSize):
              1. compute SHA-256 of chunk
@@ -355,7 +355,7 @@ sweeper needs to know not just *that* the RC reached zero but
    would drive it to zero, the *same* txn additionally writes
    `!s3|chunkblob-gc-queue|<commitTS-nanos>|<SHA>` → empty. The
    commitTS-prefixed key is the time signal: the queue is
-   naturally sorted by elegibility-start time, and any node can
+   naturally sorted by eligibility-start time, and any node can
    determine the grace-period boundary by scanning the queue with
    `endKey = !s3|chunkblob-gc-queue|<now-gracePeriod>|`. If a
    subsequent txn re-references the same SHA before the sweeper