go/types: spurious diagnostics for untyped shift operands with GoVersion < go1.13

[findleyr]

As reported in #51947 (comment), when using the new types.Config.GoVersion field with a Go version < 1.13, go/types produces inaccurate errors for untyped shift operands, as in the following code from go-cmp:

type resultFlags uint

const (
	_ resultFlags = (1 << iota) / 2

	reportEqual
	reportUnequal
	reportByIgnore
	reportByMethod
	reportByFunc
	reportByCycle
)

This is a new regression, only reachable via use of the GoVersion field, but it newly affects gopls because we are now setting the GoVersion field based on go.mod, and go-cmp still has go1.11 in their go.mod.

The fix looks straightforward but shifts are notoriously tricky and have been a source of a significant number of bugs in the past. Nevertheless, we should endeavor to fix this regression for Go 1.18.1.

CC @griesemer

[findleyr]

@gopherbot please backport to 1.18, this is a newly visible bug with no workaround.

[gopherbot]

Backport issue(s) opened: #52032 (for 1.18).

Remember to create the cherry-pick CL(s) as soon as the patch is submitted to master, according to https://go.dev/wiki/MinorReleases.

[gopherbot]

Change https://go.dev/cl/396775 mentions this issue: go/types: don't report errors for untyped int shifts on Go < 1.13