Duplicated IOActivity-to-histogram mapping: GetExternalTableFileReadHistogram is a copy of GetFileReadHistograms #14935

Expected behavior

External table reads and regular SST reads map Env::IOActivity to the FILE_READ_* histograms through one shared helper, so the two paths cannot drift apart.

Actual behavior

The mapping exists twice in the tree:

• GetFileReadHistograms — file/random_access_file_reader.cc:26 (file-local inline, not declared in any header)
• GetExternalTableFileReadHistogram — table/external_table.cc:26 (anonymous namespace), added by Wire external table reads into RocksDB metrics #14839 ("Wire external table reads into RocksDB metrics")

The two bodies are identical except for the signature line — presumably copied because the original is inline in a .cc and cannot be reused from another translation unit:

-inline Histograms GetFileReadHistograms(Statistics* stats,
-                                        Env::IOActivity io_activity) {
+Histograms GetExternalTableFileReadHistogram(Statistics* stats,
+                                             Env::IOActivity io_activity) {
   switch (io_activity) {
     ... (remaining 29 lines byte-identical) ...

Any future IOActivity case or stats-level condition added to one copy will silently miss the other: the missed path falls through to HISTOGRAM_ENUM_MAX and records nothing, so external-table reads would stop matching regular SST reads for that activity — the exact skew #14839 set out to prevent.

Steps to reproduce the behavior

Not a runtime bug yet; the copies are still in sync as of 5b1c3b3. Filing before they diverge.

Suggested fix: hoist GetFileReadHistograms into file/random_access_file_reader.h and call it from table/external_table.cc, deleting the copy. Happy to send a PR if that sounds right.