[improve][docs] Add the seek by timestamp benchmark of offloaded topics to the release highlights

[KannarFr]

Adds a benchmark section to the release highlights for searches by timestamp on offloaded topics, as discussed with @lhotari.

The new section, "Seek by timestamp in offloaded topics with fewer object store requests", follows the structure of the tailing-reads benchmark above it: the use case, the setup, a results table, what each change contributes, a chart, how it was measured, and what the results do not show.

It covers three changes to how offloaded ledgers are read:

• the reuse of entry offsets found while scanning a data block, contributed by @omarkj;
• the search reading the first entry of data blocks;
• the offsets remembered for each open offloaded ledger.

Ranged reads sent to the object store per search, with 64 MiB data blocks and a 1 MiB read buffer:

	Before	After the first change	After the second	5.0.0
Cold search, mean of 16	428	156	61	44
Warm search, mean of 16	223	14	9	9

Notes for the review:

• The numbers come from OffloadedLedgerFindPositionTest run against an in-memory object store, so they count requests and bytes, not latency. The page says so.
• "Before" is the code without any of the three changes, as in 4.0.13 and 4.2.4. The page states that the first change is also in the 4.0 and 4.2 maintenance branches for their next releases.
• The chart is static/assets/release-highlights-5.0/seek-by-timestamp-offloaded-ranged-reads.svg, with the fonts, colors and layout of the two existing charts in that directory.

The same section is added to docs/ and to versioned_docs/version-5.0.x/.

✅ Contribution Checklist

• I read the contribution guide
• I updated the versioned docs

🤖 Generated with Claude Code