Switch to eager asof join to speed up arrow queries

[jleibs]

The previous join operation was quite expensive, asof_join is much less flexible, but significantly more performant.

Before:

obj_mono_points/insert  time:   [107.87 ms 108.56 ms 109.41 ms]
                        thrpt:  [91.400 Kelem/s 92.118 Kelem/s 92.700 Kelem/s]

obj_mono_points/query   time:   [21.120 ms 21.316 ms 21.533 ms]
                        thrpt:  [4.6440 Kelem/s 4.6914 Kelem/s 4.7348 Kelem/s]

obj_batch_points/insert time:   [492.58 µs 493.57 µs 494.80 µs]
                        thrpt:  [20.210 Melem/s 20.261 Melem/s 20.301 Melem/s]

obj_batch_points/query  time:   [285.54 µs 290.11 µs 295.94 µs]
                        thrpt:  [337.91 Kelem/s 344.70 Kelem/s 350.22 Kelem/s]

After:

obj_mono_points/insert  time:   [107.06 ms 107.28 ms 107.52 ms]
                        thrpt:  [93.006 Kelem/s 93.212 Kelem/s 93.407 Kelem/s]
                 change:
                        time:   [-1.9755% -1.1735% -0.4966%] (p = 0.00 < 0.05)
                        thrpt:  [+0.4991% +1.1875% +2.0153%]
                        Change within noise threshold.

obj_mono_points/query   time:   [403.77 µs 406.58 µs 410.60 µs]
                        thrpt:  [243.55 Kelem/s 245.95 Kelem/s 247.67 Kelem/s]
                 change:
                        time:   [-98.111% -98.088% -98.064%] (p = 0.00 < 0.05)
                        thrpt:  [+5065.6% +5130.8% +5194.2%]
                        Performance has improved.

obj_batch_points/insert time:   [496.75 µs 498.94 µs 502.60 µs]
                        thrpt:  [19.897 Melem/s 20.043 Melem/s 20.131 Melem/s]
                 change:
                        time:   [+0.4017% +0.9136% +1.5290%] (p = 0.00 < 0.05)
                        thrpt:  [-1.5060% -0.9053% -0.4001%]
                        Change within noise threshold.

obj_batch_points/query  time:   [6.5295 µs 6.5356 µs 6.5426 µs]
                        thrpt:  [15.285 Melem/s 15.301 Melem/s 15.315 Melem/s]
                 change:
                        time:   [-97.699% -97.638% -97.578%] (p = 0.00 < 0.05)
                        thrpt:  [+4028.7% +4133.6% +4245.8%]
                        Performance has improved.

Checklist

• I have read and agree to Contributor Guide and the Code of Conduct
• I've included a screenshot or gif (if applicable)
• I've added a line to CHANGELOG.md (if this is a big enough change to warrant it)

[emilk]

Code LGTM, and if it passes the tests I guess it works :)

[emilk]

multiple rows per component? I don't quite understand the context for the question.

adding a debug_assert! is a good start to answering that question

[jleibs]

This call & the question both go away with #590

[emilk]

Suggested change

	Ok(df.sort([Instance::NAME], false)?)
	let reverse = false;
	Ok(df.sort([Instance::NAME], reverse)?)