Convert offsets into subtraction in RunEnd and Patches

[robert3005]

Offsets were required because we didn't have a good system for optimising and
reducing arrays, now that we do we only need to keep them in fastlanes arrays

[robert3005]

This is not ready

[codspeed-hq]

Merging this PR will degrade performance by 28.39%

❌ 13 regressed benchmarks
✅ 1003 untouched benchmarks
⏩ 1522 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	take_search_chunked[(0.005, 0.05)]	228.2 µs	311.1 µs	-26.65%
❌	Simulation	take_search_chunked[(0.005, 0.5)]	2.1 ms	2.9 ms	-28.29%
❌	Simulation	take_search_chunked[(0.005, 0.1)]	435.4 µs	601 µs	-27.56%
❌	Simulation	take_search_chunked[(0.005, 1.0)]	4.2 ms	5.8 ms	-28.39%
❌	Simulation	take_search_chunked[(0.01, 0.5)]	2.3 ms	3.1 ms	-26.74%
❌	Simulation	take_search_chunked[(0.01, 0.05)]	245.7 µs	328.6 µs	-25.22%
❌	Simulation	take_search_chunked[(0.1, 0.1)]	618.5 µs	784.4 µs	-21.15%
❌	Simulation	take_search_chunked[(0.01, 0.1)]	469.9 µs	635.6 µs	-26.07%
❌	Simulation	take_search_chunked[(0.1, 0.05)]	330.8 µs	413.8 µs	-20.06%
❌	Simulation	take_search_chunked[(0.1, 1.0)]	5.8 ms	7.5 ms	-22.15%
❌	Simulation	take_search_chunked[(0.1, 0.5)]	2.9 ms	3.8 ms	-22.02%
❌	Simulation	take_search_chunked[(0.01, 1.0)]	4.5 ms	6.2 ms	-26.84%
❌	Simulation	map_each[BufferMut<i32>, 128]	770.6 ns	858.1 ns	-10.2%

---

_{Comparing rk/nooffsets (cb967cf) with develop (f03a0c7)}

Footnotes

1. 1522 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩