Put long multiline constrained SRTP on separate lines.

[nojaf]

Could you please review @dsyme if this is in line with the style guide?
I do notice that Fantomas doesn't put a space before the colon in when ^T :.
Is this by design for constraints?

Fixes #2230.

[dsyme]

I do notice that Fantomas doesn't put a space before the colon in when ^T :.
Is this by design for constraints?

This is ok

[dsyme]

Historically I would have indented this further considering the when as part of the type , rather than part of the declaration. I'm not sure if this will cause problems. For example it could be like this:

val inline average:
    array: ^T[] -> ^T
        when ^T: (static member (+): ^T * ^T -> ^T)
        and ^T: (static member DivideByInt: ^T * int -> ^T)
        and ^T: (static member Zero: ^T)

I suppose this PR should add some examples of multi-line types (curried, tupled arguments) to check things are ok?

For example this sort of thing - https://fsprojects.github.io/fantomas-tools/#/fantomas/preview?data=N4KABGBEDGD2AmBTSAuKBtAPAYVgWwAcBLAG0XgDkBDPRACgB1IBBAN0QCcqBzRJgSgB8AXQYcGAO1ZUSYIhJLzEYKuy680VDlwCeaAHoAVdMLABaQSu1U9YIyfOWtug8dMWrLu28eebrhw9nf29Ap2tbe3dwryjfYMifDyMxSQgIAHcAC0QJbzQ6AGcAFypiomgwWjwAI04wOgBqflcwACpvXyN%2BVLz0qgl4fIaSsoqqxFr6gBEiViIkACEdAEkJYtaO%2BWKuwx7xPogBoaMC0fLK6rqOMAAtTlhXfkgAGnAoQsRi8oluQtQwOh3hBQOl0pAACTFHQEZBoSDbV7A8EQ6QkACucMByLBAAY3mCwZA1kh1gBlIgAL2QBMJEAALDiwMJkQBfWkgpmQ6GwgEI9ZIumQtGYgFAukQACMHMJkAAslQAB4AGSUyty3GKWUFEslACZcUyWWD2cjQbKoTCsZBcvAAPIAM1VEkQZOhZB1RNRMlFaHFEr1MqJAFFBo7nTSmRBICQHZAjWyZeavTzrTVYLASJ6USKsf66QBmIPgtafDjFABi8hkFEQGUULuzYOKHExCZNSa5lt58PTmabUG9GLzUbA9OL0bJBCo0EQi0QDtgHEQAAUtDQvpwBxAW226cb0qawcmUam%2BX2s8XhT6RxKwABWCdQKczucLpeIZWwDKcaBUT5rKwsB-uUsASNuYC7og7aHp2QrdmmGaXl2uZiqOABsT6QC%2Bs7zouy4AKoELCHB-gBUjAWMYEQQ6MifDBEBHukJ7Rgh55IQO17Dmhd4AOxYThb74Yg2AkP%2BhS4BIJSttAxRLjRdHQfuiZml2Z69hxV5Dr62J3gAHAJ064e%2By5ypM1wKSQ9HKR2qnwepUAXpx2m3hKACchmvnhH64CQ1FPrRVlKYSB6MXBFoOZATlaahfqjpK%2BKjthRmIMwDrFJwuB4HgVAQVBDFgExnL2Va7H9jFN48bq0pJYJ3nLmSkwVJm-mjoF1khSpx5qaVGnlShlVxXe%2BqebOaUZRwjV4M1fngU%2B%2BU2bBdkRb1jmaQN3FDbqRa1SlzAcLA6KDNMiCKNNE15a2wVgqFhXhSmq38sUzmxbpurjklCqKisDqGDkEjBkFZJZEuxQAOoLFqEH0oai1hctD09lAiIVZtb10pKj6fUqawOkQip2iRZRLsGioEMuhSFEQrV3vesOdbZ3UlUjT0vYN6OEpKmHY4qABKiBwBw8AQ-AUNPjDBVFWALGDpFKMbTp%2Bac-xPP84LlDolMHCOisGV4P8T6SpL92no9eDoiQ5QNogFZLjl3ycIYpWo4r8UGUlatLvAcoW1bSi2xw9sXVh0BZOusmcAA%2BhkkPasbCOmyz8vwa9StgpKHk8-tuh2hwqolCLYujhLcN3QnrFywKLuuXSBpYV92c2Ln%2BfFBQmvXDresG-F8dMytLPm5bpD%2B3bZQTU7vLV1Vtc1XeLARM3RAlD7Q-WwHQdbiHYdcBHHDR7H8al1LMvco9ycWqno56oGPMAGo3os8jwPI3CF9qT56fTN1dcxPVJ1XCsa6Ej1DtOeX0KxHVktTCQj9BgvzftDL%2B6RbrHz-tac%2BXpL53j1B9MBSppiwGKAAcS%2BKTcmiBKbQIQU%2BOmvdf7M0Qv1FO7M07pD1FjMBvth4ukWH5aAABrRYO9%2BFfEKHaCQZINy%2BU1nNNqik6HFX7ow5CzC0asIgHqbmc9az1iUPOYoP5cgT0QCdPGEgiCgQkMwQYZktbdzvO1a6yCf6KMRsotmair4qznswRQ3AJAQIkFAsCFJ-FlHRMuQwsASS5FKJYyyHVv6M3oUosqKiL4sKvu7HxltOASDGOwL8vxbHXFMfICx0D7ESkcQo6WaC0keNdtgzOnCh4RmVDQGo8AqCiVgFTX4OjrYJKcRAFBJsK6PWioA6ehICyJTnmQzgRBaDrBkAAaUQIgAgMT1hrCEQMUOwzaknzYn1dJmDMl3gLLPCUkBeEDH4RGQo%2B1DqDFrCUcgK8-YujIRTKmYEql0gWgzJafc3ENKnltQsN856LC0PVExS9oAcGWdWDK8ACLmLAidaAYkuDxICvIo%2B4zZaTPWqoppEoCygNuYslFKzSgkDdAddE0l0QEDdDoD0hKgrHPqfCDBOZLlUtwbcjZWyvpty1o6e5EhHlKEBZzXESDRkuLqQwiF0yoWzI4bc5lFRihygQJGBxRKQWMXeMIWkCJCgVipgCfKrIgA

[nojaf]

Hmm, yes looking at https://docs.microsoft.com/en-us/dotnet/fsharp/style-guide/formatting#formatting-types-in-signatures

val inline average: 
    array: ^T[] -> ^T 
        when ^T: (static member (+): ^T * ^T -> ^T) 
        and ^T: (static member DivideByInt: ^T * int -> ^T) 
        and ^T: (static member Zero: ^T)

does make more sense.

[nojaf]

@dsyme, I've added some more tests.
Please verify this last commit if this is good to go.

[dsyme]

Looks good!!