Age-ordered RS

[tilk]

This PR changes RS so that oldest entries are taken first. The idea is that the indexes in get_ready_list and take are no longer physical RS row numbers, but they are ordered from oldest to newest. The priority encoder in WakeupSelect then naturally selects the oldest ready entry.

The main RS test was extended to introduce random delays and testing multiple ready lists. The individual method tests are dropped, for three reasons:

1. These tests don't just verify conformance to some protocol, they need to look into RS internals to work. Because of that, changes to internals invalidate these tests, creating more work for the developer.
2. There is quite a lot of lines of code in these tests, which increases the amount of maintenance work even more.
3. The main test seems good enough, and even if some error is not found there, the tests for the whole core will most probably fail then.

[github-actions]

Benchmarks summary

Performance benchmarks

aha-mont64	crc32	minver	nettle-sha256	nsichneu	slre	statemate	ud
▲ 0.421 (+0.004)	▲ 0.539 (+0.014)	▲ 0.373 (+0.003)	▲ 0.632 (+0.001)	▲ 0.359 (+0.000)	▲ 0.293 (+0.002)	▼ 0.327 (-0.000)	▲ 0.442 (+0.002)

You can view all the metrics here.

Synthesis benchmarks (basic)

Device utilisation: (ECP5)	LUTs used as DFF: (ECP5)	LUTs used as carry: (ECP5)	LUTs used as ram: (ECP5)	Max clock frequency (Fmax)
▼ 13274 (-209)	▲ 4407 (+9)	1456 (0)	1164 (0)	▲ 50 (+1)

Synthesis benchmarks (full)

Device utilisation: (ECP5)	LUTs used as DFF: (ECP5)	LUTs used as carry: (ECP5)	LUTs used as ram: (ECP5)	Max clock frequency (Fmax)
▼ 21961 (-3858)	▲ 7022 (+9)	1818 (0)	1216 (0)	▲ 41 (+1)