fix(asap-tools): send ClickHouse ticks at once in the paced feeder

[zzylol]

Summary

Follow-up to #793. The review fix there made each tick send its sub-second slots one at a time, waiting for every request of a slot before sending the next, so ASAPQuery's per-group watermark never sees a group's samples out of order. ClickHouse doesn't need that ordering (MergeTree sorts on insert), and the waiting roughly doubled its tick time at 2 samples/s: about 450 ms per tick before #793's review fix, about 950 ms after it, at 1e6 rows/s.

Now only remote write sends slot by slot; ClickHouse gets the whole tick at once, as before the review fix.

Testing

On a CloudLab c6320, against ClickHouse 26.10 and precompute_engine from main:

Check	Result
ClickHouse, 1e6 rows/s, 2 samples/s, 10 s, feeder pinned to 2 cores	0 overrun ticks, ticks 356–617 ms (about 860–1030 ms without this change)
ClickHouse, 1e4 series × 4 samples/s, 10 s	400,000 rows, unique (series, ts), 40 rows per series at 250 ms spacing; every value matches an independent Python reimplementation of the generator exactly
Remote write, same workload, engine with --allowed-lateness-ms 0 and a Sum by label_0 1 s tumbling aggregation	0 late drops (the feeder before #793's review fix: 236,000 of 400,000 dropped), 4,000 group batches of 100 samples, 1,000 windows emitted
Remote write, 1e6 rows/s at 1 and 2 samples/s, feeder pinned to 2 cores	0 overrun ticks, 0 failed requests

cargo clippy --all-targets -- -D warnings, cargo fmt --check and the 6 unit tests pass.

🤖 Generated with Claude Code

https://claude.ai/code/session_01AnJ7LocWxguh3q4QWkCYae