ci: add Go 1.27

[jsha]

No description provided.

[jsha]

Test failures:

--- FAIL: TestConfigValidation (0.04s)
    --- FAIL: TestConfigValidation/mtca.json (0.00s)
        main_test.go:72: Failed to validate config file "mtca.json": open ../../test/config/mtca.json: no such file or directory
FAIL

This is because TestConfigValidation checks cmd.AvailableConfigValidators(), and the mtca component is gated on //go:build go1.27, and mtca.json is only available for the plain config/ directory. That worked out fine before because we used gotip for all config-next build phases, but now we need to have a copy of mtca.json for config/.

--- FAIL: TestPublish (0.06s)
    mtpublisher_test.go:200: selecting the latest checkpoint: failed to select one checkpoints: sql: no rows in result set
FAIL

I believe this is a race between the mtca and mtpublisher unittests, the same as the race between the ra and sa ones. Not sure why this is only showing up under Go 1.27; perhaps we got lucky before.

[aarongable]

I believe this is a race between the mtca and mtpublisher unittests, the same as the race between the ra and sa ones. Not sure why this is only showing up under Go 1.27; perhaps we got lucky before.

Is there any chance we can avoid repeating our RA unittest mistakes in the MTCA and MTPub? Those two components interact with very few database tables; can we just fake those in memory?

[jsha]

Is there any chance we can avoid repeating our RA unittest mistakes in the MTCA and MTPub? Those two components interact with very few database tables; can we just fake those in memory?

Yep, I think we should do better here, as a separate PR.

The golangci-lint failures are real, because we need to update to a golangci version with Go 1.27 support. Doing that produces a bunch of new findings; I'm working on fixing those.