fix: Paths::primary() is now what used to be Paths::primary_fallible()

[larseggert]

And deal with the fallout.

Fixes #1828
Fixes #1813

[github-actions]

Benchmark results

Performance differences relative to d74ad5e.

• drain a timer quickly time: [427.80 ns 434.15 ns 440.13 ns]
  change: [+37.484% +40.281% +42.933%] (p = 0.00 < 0.05)
  💔 Performance has regressed.

• coalesce_acked_from_zero 1+1 entries
  time: [198.94 ns 199.53 ns 200.15 ns]
  change: [-0.2568% +0.1136% +0.4518%] (p = 0.54 > 0.05)
  No change in performance detected.

• coalesce_acked_from_zero 3+1 entries
  time: [242.70 ns 243.31 ns 243.96 ns]
  change: [+0.2622% +0.6440% +1.0581%] (p = 0.00 < 0.05)
  Change within noise threshold.

• coalesce_acked_from_zero 10+1 entries
  time: [242.14 ns 242.96 ns 243.94 ns]
  change: [+0.6409% +1.2215% +1.7800%] (p = 0.00 < 0.05)
  Change within noise threshold.

• coalesce_acked_from_zero 1000+1 entries
  time: [226.71 ns 226.91 ns 227.14 ns]
  change: [+2.9448% +3.7346% +4.5235%] (p = 0.00 < 0.05)
  💔 Performance has regressed.

• RxStreamOrderer::inbound_frame()
  time: [118.66 ms 118.72 ms 118.79 ms]
  change: [-0.5968% -0.5277% -0.4565%] (p = 0.00 < 0.05)
  Change within noise threshold.

• transfer/Run multiple transfers with varying seeds
  time: [119.33 ms 119.64 ms 119.94 ms]
  thrpt: [33.350 MiB/s 33.434 MiB/s 33.520 MiB/s]
  change:
  time: [-2.6705% -2.3074% -1.9567%] (p = 0.00 < 0.05)
  thrpt: [+1.9957% +2.3619% +2.7438%]
  Change within noise threshold.

• transfer/Run multiple transfers with the same seed
  time: [119.76 ms 119.95 ms 120.15 ms]
  thrpt: [33.291 MiB/s 33.346 MiB/s 33.400 MiB/s]
  change:
  time: [-2.3861% -2.1574% -1.9302%] (p = 0.00 < 0.05)
  thrpt: [+1.9682% +2.2049% +2.4444%]
  Change within noise threshold.

• 1-conn/1-100mb-resp (aka. Download)/client
  time: [1.0684 s 1.0914 s 1.1237 s]
  thrpt: [88.993 MiB/s 91.621 MiB/s 93.594 MiB/s]
  change:
  time: [-7.0514% -3.5731% -0.2377%] (p = 0.07 > 0.05)
  thrpt: [+0.2382% +3.7055% +7.5863%]
  No change in performance detected.

• 1-conn/10_000-parallel-1b-resp (aka. RPS)/client
  time: [424.94 ms 427.15 ms 429.36 ms]
  thrpt: [23.291 Kelem/s 23.411 Kelem/s 23.533 Kelem/s]
  change:
  time: [-0.2948% +0.4128% +1.1808%] (p = 0.28 > 0.05)
  thrpt: [-1.1670% -0.4111% +0.2957%]
  No change in performance detected.

• 1-conn/1-1b-resp (aka. HPS)/client
  time: [50.247 ms 50.595 ms 50.956 ms]
  thrpt: [19.625 elem/s 19.765 elem/s 19.902 elem/s]
  change:
  time: [-1.7640% -0.7892% +0.1775%] (p = 0.12 > 0.05)
  thrpt: [-0.1772% +0.7954% +1.7956%]
  No change in performance detected.

Client/server transfer results

Transfer of 134217728 bytes over loopback.

Client	Server	CC	Pacing	Mean [ms]	Min [ms]	Max [ms]	Relative
msquic	msquic			704.7 ± 317.2	374.5	1240.2	1.00
neqo	msquic	reno	on	904.7 ± 180.5	759.1	1274.4	1.00
neqo	msquic	reno		1032.1 ± 234.3	784.6	1412.2	1.00
neqo	msquic	cubic	on	1016.0 ± 248.8	789.7	1376.9	1.00
neqo	msquic	cubic		1050.1 ± 217.6	775.0	1337.0	1.00
msquic	neqo	reno	on	4573.9 ± 312.5	4227.2	5135.3	1.00
msquic	neqo	reno		4499.3 ± 244.6	4206.4	4969.0	1.00
msquic	neqo	cubic	on	4541.2 ± 196.1	4234.6	4799.1	1.00
msquic	neqo	cubic		4399.7 ± 130.6	4225.8	4619.2	1.00
neqo	neqo	reno	on	4018.6 ± 220.2	3681.5	4404.7	1.00
neqo	neqo	reno		3575.4 ± 202.5	3240.4	3802.7	1.00
neqo	neqo	cubic	on	4319.4 ± 287.9	3668.4	4793.2	1.00
neqo	neqo	cubic		4295.7 ± 300.5	3794.7	4947.0	1.00

⬇️ Download logs

[martinthomson]

Some suggestions...