fix: gate read-only Graph Drop and flush against on-disk writes

[shivamka1]

Summary

Read-only Graph.load(path, read_only=True) handles were writing to the
graph directory on drop, corrupting a live writer's crash recovery:

• Bug fixed paper typo, centered images #1: Drop for GraphStore appended a shutdown checkpoint to the
  shared WAL at LSN 256, clobbering the writer's live records. On next
  reopen after a writer crash, recovery failed with
  Expected checkpoint at given LSN.
• Bug graph entities clean up, refactor, formatted with scalafmt #2: Drop for Storage rewrote .meta even when the reader's
  snapshot was stale, rolling the cached counts backwards.

Fix: add a read_only: bool field to GraphStore and Storage, set by
load_read_only, and short-circuit both Drop impls (plus flush())
when the flag is true. Read-only handles are now byte-passive with
respect to the graph directory.

• GraphStore gets load_read_only + Drop/flush guards
• Storage gets a matching flag + Drop guard for the .meta refresh
• TemporalGraph::load_inner threads read_only=true through to
  Layer::load_read_only
• New StorageError::ReadOnly variant, returned by flush() on a
  read-only handle
• In-process Storage::read_only() also sets the flag so it doesn't
  double-refresh .meta alongside the original writer

Bug #3 (reader Graph.load racing writer segment-file creation, ENOENT
at disk_layer_stats/mod.rs:301) is a separate, unrelated race in the
reader-open path; it's documented and covered by an xfail regression
test in this PR so a future fix will trip the sentinel.

Test plan

New file: python/tests/test_base_install/test_graphql/test_read_only_durability.py

• CONTROL — writer + SIGKILL with no readers recovers cleanly
• Bugs fixed paper typo, centered images #1/graph entities clean up, refactor, formatted with scalafmt #2 — SIGSTOP the writer, snapshot every file in the graph
  dir, open+drop a read-only handle, snapshot again; assert
  byte-identical. Fails on main, passes on this branch.
• Bug Add Snapshotting features #3 (xfail) — 4 reader threads + live writer; asserts no
  disk_layer_stats/mod.rs ENOENT. Still fails; xfail tracker for
  the follow-up.
• All 11 existing read-only tests
  (test_read_only_load.py, test_read_only_in_memory.py) pass
  unchanged.

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Rust Benchmark'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 2.

Benchmark suite	Current: 6e3c10a	Previous: 9823ef7	Ratio
lotr_graph/num_edges	4 ns/iter (± 0)	0 ns/iter (± 0)	+∞
lotr_graph/num_nodes	5 ns/iter (± 0)	1 ns/iter (± 0)	5
lotr_graph/graph_latest	3 ns/iter (± 0)	0 ns/iter (± 0)	+∞
lotr_graph_materialise/materialize	7800122 ns/iter (± 38894)	1564816 ns/iter (± 35303)	4.98
lotr_graph_window_100/num_nodes	14 ns/iter (± 0)	5 ns/iter (± 0)	2.80
lotr_graph_window_100_materialise/materialize	7817239 ns/iter (± 34090)	1669150 ns/iter (± 10700)	4.68
lotr_graph_window_10/has_node_existing	136 ns/iter (± 9)	62 ns/iter (± 11)	2.19
lotr_graph_window_10_materialise/materialize	3304261 ns/iter (± 16881)	971980 ns/iter (± 4278)	3.40
lotr_graph_subgraph_10pc_materialise/materialize	2037283 ns/iter (± 22219)	334634 ns/iter (± 1287)	6.09
lotr_graph_subgraph_10pc_windowed/has_node_existing	142 ns/iter (± 10)	62 ns/iter (± 14)	2.29
lotr_graph_subgraph_10pc_windowed_materialise/materialize	1251820 ns/iter (± 9826)	230399 ns/iter (± 2617)	5.43
lotr_graph_window_50_layered/num_edges_temporal	145011 ns/iter (± 2076)	70121 ns/iter (± 7586)	2.07
lotr_graph_window_50_layered/has_node_existing	391 ns/iter (± 21)	129 ns/iter (± 12)	3.03
lotr_graph_window_50_layered/graph_latest	83106 ns/iter (± 1880)	36649 ns/iter (± 916)	2.27
lotr_graph_window_50_layered_materialise/materialize	28766145 ns/iter (± 68680)	3488825 ns/iter (± 24948)	8.25
lotr_graph_persistent_window_50_layered/num_edges_temporal	582198 ns/iter (± 4592)	192686 ns/iter (± 1569)	3.02
lotr_graph_persistent_window_50_layered/has_node_existing	426 ns/iter (± 384)	174 ns/iter (± 83)	2.45
lotr_graph_persistent_window_50_layered/graph_latest	122169 ns/iter (± 1992)	57549 ns/iter (± 4809)	2.12
lotr_graph_persistent_window_50_layered_materialise/materialize	49154804 ns/iter (± 297824)	5298035 ns/iter (± 147912)	9.28

This comment was automatically generated by workflow using github-action-benchmark.