infinitesimal non-zero float values are rejected by the preprocessor

[thehowl]

this comes from porting package math. Reproducible example:

package main

const (
	SmallestNonzeroFloat64 = 0x1p-1022 * 0x1p-52            // 4.9406564584124654417656879286822137236505980e-324
	DividedByTwo           = SmallestNonzeroFloat64 / 2
)

func main() {
	println(DividedByTwo)
}

In one of the tests, the following expression appears:

SmallestNonzeroFloat64 / 2

SmallestNonzeroFloat64 is what it says it is (precisely: 0x1p-1022 * 0x1p-52). Of course, being already the smallest, dividing it by 2 should yield 0. However, at preprocessing stage, the Gnovm rejects it entirely:

panic: cannot convert untyped bigdec to float64 -- too close to zero [recovered]
        panic: gno.land/r/GsgkK3Vq/all_test.gno:3282: cannot convert untyped bigdec to float64 -- too close to zero

goroutine 1 [running]:
github.com/gnolang/gno/gnovm/pkg/gnolang.Preprocess.func2.1()
        /home/howl/oc/gno2/gnovm/pkg/gnolang/preprocess.go:173 +0x46d
panic({0xbeb9a0, 0xf45fb0})
        /usr/lib/go/src/runtime/panic.go:890 +0x263
github.com/gnolang/gno/gnovm/pkg/gnolang.ConvertUntypedBigdecTo(0xc0084fd8f0, {0xc00b4797a8?}, {0xf521c8, 0xf44fb0})
        /home/howl/oc/gno2/gnovm/pkg/gnolang/values_conversions.go:1258 +0x585