cmd/go: TestScript/mod_load_badchain flake on windows-amd64-longtest

[bcmills]

2019-11-12T05:35:33-99957b6/windows-amd64-longtest

In CL 204777 I made the "finding versions" log message nondeterministic based on the actual delay of the operation. Perhaps it should be made deterministic — or at least made deterministic during testing — so that output during tests doesn't depend on the load factor of the test machine.

CC @jayconrod

--- FAIL: TestScript (0.00s)
    --- FAIL: TestScript/mod_load_badchain (15.88s)
        script_test.go:192: 
            # Download everything to avoid "finding" messages in stderr later. (7.343s)
        […]
            > cmp stderr list-missing-test-expected
            [diff -stderr +list-missing-test-expected]
            -go: finding versions for example.com
             go: found example.com/badchain/c in example.com/badchain/c v1.1.0
             go: m/testuse tested by
             	m/testuse.test imports
             	example.com/badchain/c: example.com/badchain/c@v1.1.0: parsing go.mod:
             	module declares its path as: badchain.example.com/c
             	        but was required as: example.com/badchain/c
            
            FAIL: testdata\script\mod_load_badchain.txt:35: stderr and list-missing-test-expected differ
            
FAIL
FAIL	cmd/go	376.850s

[gopherbot]

Change https://golang.org/cl/213679 mentions this issue: cmd/go: make "finding" logging deterministic

[bcmills]

We should make the logging deterministic before we cut the RC.

The nondeterministic logging is too confusing (see also #36258).