AVM: Extract divideCeilUnsafely to help document opcode costing rationale

[michaeldiamant]

Extracts divideCeilUnsafely function to help document opcode costing rationale. Stems from discussion in #3857 (comment).

Analysis below shows divideCellUnsafely is inlined. Implies there's no overhead for adding the function. The logs show divideCell because I renamed the method after an initial pass.

~/dev/go-algorand/data/transactions/logic divide_ceil !1 ?2 ───────────────────────────────────────────────────────────────────── 04:06:17 PM
❯ go build -gcflags=-m=2 . 2>&1  | grep opcodes.go | grep 'divideCeil'
./opcodes.go:73:6: can inline divideCeil with cost 8 as: func(int, int) int { return (numerator + denominator - 1) / denominator }
./opcodes.go:77:6: can inline (*linearCost).compute with cost 43 as: method(*linearCost) func([]stackValue) int { cost := lc.baseCost; if lc.chunkCost != 0 && lc.chunkSize != 0 { cost += divideCeil(lc.chunkCost * len(stack[len(stack) - 1].Bytes), lc.chunkSize) }; return cost }
./opcodes.go:81:21: inlining call to divideCeil func(int, int) int { return (numerator + denominator - 1) / denominator }
./opcodes.go:140:28: inlining call to (*linearCost).compute method(*linearCost) func([]stackValue) int { cost := lc.baseCost; if lc.chunkCost != 0 && lc.chunkSize != 0 { cost += divideCeil(lc.chunkCost * len(stack[len(stack) - 1].Bytes), lc.chunkSize) }; return cost }
./opcodes.go:140:28: inlining call to divideCeil func(int, int) int { return (numerator + denominator - 1) / denominator }

As intended, benchmarking shows no meaningful change.

• Benchmark run: go test ./data/transactions/logic -bench=BenchmarkUintMath
• BenchmarkUintMath-diff.txt - old = master and new = the PR.

[jannotti]

Yes, this is what I was thinking. I probably wouldn't say 'Unsafely' as that seems to be what division always is. But not something I feel strongly about.