Deprecate assert_display_snapshot, refactor macros

[max-sixty]

This PR:

• Deprecates assert_display_snapshot in favor of assert_snapshot, as discussed in assert_display_snapshot vs. assert_snapshot #379
• Makes a sizable change in how macros are constructed:
  • It makes macros consistent — previously debug & display had different forms than the others, not taking a debug_expr.
  • The non-redaction macros are a simple wrapper of _assert_snapshot_base, which takes a closure transform of how to transform the input to a string.
  • The redaction macros are a simple wrapper of _assert_serialized_snapshot, which takes an arg format (similar to before), and uses that to construct a transform.
  • As a result, the boilerplate for each macros is removed.

It's not quite complete, but wanted to post before spending more time on it. Now complete

Previous todos:

• We get a couple of test failures; I need to look into this; we seem to be generating an additional --- in some yaml snapshots. (possibly something is being converted to string twice?)
• Update docs; previously the debug_expr wasn't well documented, and now that the source is more opaque this is more important

[max-sixty]

This is updated, with all the tests passing, and better comments around the macros.

One thing that would make the code even simpler — serializing values the same for File and Inline. I think the only difference is that YAML snapshots lose the leading --- in files — is that right? Any reason not to remove for both files & inline?

[max-sixty]

Updated the docs at mitsuhiko/insta-website#17

[mitsuhiko]

I will try to get this reviewed this week.

[max-sixty]

@mitsuhiko the gentlest of pings! Also will take feedback if you think it's not so great :)

[mitsuhiko]

I am in favor of this. I will try to get it merged and released ahead of backwards incompatible changes.

[max-sixty]

I am in favor of this. I will try to get it merged and released ahead of backwards incompatible changes.

Thanks!

[kennethloeffler]

Just a heads up, @mitsuhiko + @max-sixty - I have good reason to believe that these changes broke one of our tests. It seems like this version of the assert_yaml_snapshot macro takes ownership of the value, while previous versions only borrowed it. I can reproduce the test failures observed in rojo-rbx/rbx-dom#398 on insta 1.36.0 (the version in which the changes in this PR shipped), but not on insta 1.35.1.

[max-sixty]

I'll take a look, thanks @kennethloeffler !

[max-sixty]

This is fixed in #453, apologies for the temporary break

[kennethloeffler]

It's perfectly okay - it's easy to see how this flew under the radar 😅

[mitsuhiko]

Sorry for that. I pushed out a fix!