add Scala 3.0.0-RC1 to crossbuild

[SethTisue]

partially addresses #570

[SethTisue]

@dwijnand you might know what this is about...?

[error] -- Error: /home/travis/build/playframework/play-json/play-json/shared/src/test/scala/play/api/libs/json/MacroSpec.scala:106:52 
[error] 106 |      implicit val familyReads: Reads[Family]     = Json.reads[Family]
[error]     |                                                    ^^^^^^^^^^^^^^^^^^
[error]     |cannot reduce summonFrom with
[error]     | patterns :  case t @ 
[error]     |  _:
[error]     |    
[error]     |      play.api.libs.json.Reads[play.api.libs.json.MacroSpec.Lorem[Any] & 
[error]     |        play.api.libs.json.MacroSpec.Family
[error]     |      ]
[error]     |    
[error]     | This location contains code that was inlined from package.scala:144
[error]     | This location contains code that was inlined from JsMacroImpl.scala:122
[error]     | This location contains code that was inlined from JsMacroImpl.scala:81
[error]     | This location contains code that was inlined from JsMacroImpl.scala:82
[error]     | This location contains code that was inlined from JsMacroImpl.scala:52
[error]     | This location contains code that was inlined from JsMacroImpl.scala:22
[error]     | This location contains code that was inlined from JsMacros.scala:28

[mkurz]

Why M3 as well?

[SethTisue]

it can be convenient for downstream maintainers to publish for the last two versions, so that people can take the library version bump separately from the Scala version bump, to help narrow down the number of possible causes if something goes wrong

@dwijnand regardless, if keeping M3 makes any trouble for supporting RC1, it is totally fine with me to just drop M3

[dwijnand]

@dwijnand you might know what this is about...?

[error] -- Error: /home/travis/build/playframework/play-json/play-json/shared/src/test/scala/play/api/libs/json/MacroSpec.scala:106:52 
[error] 106 |      implicit val familyReads: Reads[Family]     = Json.reads[Family]
[error]     |                                                    ^^^^^^^^^^^^^^^^^^
[error]     |cannot reduce summonFrom with
[error]     | patterns :  case t @ 
[error]     |  _:
[error]     |    
[error]     |      play.api.libs.json.Reads[play.api.libs.json.MacroSpec.Lorem[Any] & 
[error]     |        play.api.libs.json.MacroSpec.Family
[error]     |      ]
[error]     |    
[error]     | This location contains code that was inlined from package.scala:144
[error]     | This location contains code that was inlined from JsMacroImpl.scala:122
[error]     | This location contains code that was inlined from JsMacroImpl.scala:81
[error]     | This location contains code that was inlined from JsMacroImpl.scala:82
[error]     | This location contains code that was inlined from JsMacroImpl.scala:52
[error]     | This location contains code that was inlined from JsMacroImpl.scala:22
[error]     | This location contains code that was inlined from JsMacros.scala:28

Lorem <: Family and there definitely is an instance of Reads[Lorem[Any]] in scope, so you'd expect that it can summon an instance of Reads[Lorem[Any] & Family] and indeed it did in M3. Is it not intentional or not? I'm 98% sure I can work around it, but is it a regression? A blocking one? 🤷🏼‍♂️

[SethTisue]

thanks for figuring this out, Dale.

this appears to be ready for merge.