Upgrade backends

[qmuntal]

This PR upgrades the three backends. This is a summary of the changes:

• The CNG nor the OpenSSL backends stopped using encoding/binary, so we can drop it from TestDependencies. Don't use encoding/binary go-crypto-winnative#90 and Don't use encoding/binary golang-fips/openssl#257.
• The OpenSSL backend now does the error handling in the C code. Move error handling to C code golang-fips/openssl#265
• The Darwin backend implementes hash functions using CryptoKit instead of CommonCrypto. Move oneshot hashes from xcrypto to cryptokit go-crypto-darwin#41 and Use cryptokit for implementing hash interfaces for MD5, SHA1,SHA256,SHA384 and SHA512 go-crypto-darwin#46.

[dagood]

Some unexpected allocations, and some that look like they're missing info in the test browser, but maybe I'm just not familiar:

=== RUN   TestGeneratedHash
    shared_test.go:89: go install -buildmode=shared -linkshared ./issue30768/issue30768lib
    shared_test.go:1125: executing /__w/1/s/go/bin/go test -linkshared ./issue30768 failed exit status 1:
--- FAIL: TestGeneratedHash (5.32s)

[qmuntal]

Some unexpected allocations

Filled microsoft/go-crypto-darwin#43.

and some that look like they're missing info in the test browser, but maybe I'm just not familiar:

Can reproduce this locally. This is the only info reported by the test, unfortunately. Sent https://go-review.googlesource.com/c/go/+/659456 to add the missing context.

[qmuntal]

Submitted golang-fips/openssl#268 to fix the AZL3 issue.

[qmuntal]

There are two outstanding failures:

• TestIssue71943: The test is testing that the compiler can stacka-locate a concatenated string passed to sha256.Sum256. It looks like it can't do that when using the OpenSSL and Darwin backends, but it can when using CNG. Will skip this test for now (and file an issue), it might require upstream changes. cmd/compile/internal/test: TestIssue71943 fails with OpenSSL and Darwin backends #1629.
• TestBoringInternalLink: still investigating. Edit: Fixed in d8d9bdd.

[dagood]

• TestBoringInternalLink: still investigating. Edit: Fixed in d8d9bdd.

Looks like it still happened: https://dev.azure.com/dnceng-public/public/_build/results?buildId=1015443&view=ms.vss-test-web.build-test-results-tab&runId=27170570&paneView=debug&resultId=155368

[qmuntal]

• TestBoringInternalLink: still investigating. Edit: Fixed in d8d9bdd.

Looks like it still happened: https://dev.azure.com/dnceng-public/public/_build/results?buildId=1015443&view=ms.vss-test-web.build-test-results-tab&runId=27170570&paneView=debug&resultId=155368

Yep, that commit fixed the Ubuntu, Mariner 2, and AZL 3 builds, but Mariner 1 is still failing with the same error. I can't reproduce it locally using the same base image, weird. Don't know if it worth investigating given #1634.

Darwin is also timing out...

[mertakman]

• TestBoringInternalLink: still investigating. Edit: Fixed in d8d9bdd.

Looks like it still happened: https://dev.azure.com/dnceng-public/public/_build/results?buildId=1015443&view=ms.vss-test-web.build-test-results-tab&runId=27170570&paneView=debug&resultId=155368

Yep, that commit fixed the Ubuntu, Mariner 2, and AZL 3 builds, but Mariner 1 is still failing with the same error. I can't reproduce it locally using the same base image, weird. Don't know if it worth investigating given #1634.

Darwin is also timing out...

Mariner 1.1 is at end of life, should we drop support to that?

[qmuntal]

Finally manage to make CI happy without skipping any test 😄 .