chore(release): 1.3.10 — ship the scoped-count cache that stops the sidebar re-walking the mailbox

[andrei-hasna]

Ships the scoped-count cache merged in #206 as 1.3.10. Version bump only — no code change in this PR.

Why this needs to ship rather than sit on main

emails ui is the client behind the 2026-07-31 runaway egress (370fd95e). The bounding fixes already in 1.3.9 cut it roughly 14x, and 1.3.9 is still expensive when a mailbox address is selected in the sidebar. Measured on the installed build's code path today, against the production serve, mailbox 175,344 messages:

BEFORE (main without #206)  357s  3 walks  618 reqs  213.1 MB  104 req/min  2.10 GB/h
AFTER  (main with #206)     334s  1 walk   206 reqs   71.0 MB   37 req/min  0.75 GB/h

walks at t = [0, 136, 263]  ->  walks at t = [0]; 10 of 11 ticks issued zero requests

3.00x in the measured window; 6.8x in steady state (206 req/132s becomes 206 req/900s). The window understates it because it contains the fixed build's one-and-only walk and none of its cached interval.

A merged fix changes nothing for the process that is actually running — the TUI runs the installed package, so this is the rung that makes the fix real.

What this does NOT fix

Each refill is still one whole walk of 206 requests and 71 MB; only the duty cycle improves. /v1/messages/counts accepts ?to= and silently discards it (measured today: ?to=<real> and ?to=<cannot-exist> both return the whole-store total 516565, while ?domain= correctly returns 361410). Removing the burst needs that server-side filter, tracked as 3ae0181e.

Refs: 41529344, 370fd95e, 3ae0181e

Agent: agent-chief-operations

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}