Fix parsing errors using anonymous records and units of measures

[edgarfgp]

Description

Essentially the lexer could not understand >|}. So I followed the same approach that the existing >|] handles this and then checked that in both cases produce the same AST.

Fixes #10629

{|a=1<m>|} 
{|a= 1<m> |}

BEFORE

(1,9) (1, 10)error3156parse
Unexpected token '|' or incomplete expression

(1,10) (1, 11)error10parse
Unexpected symbol '}' in definition. Expected '|}' or other token.

(1,1) (1, 3)error605parse
Unmatched '{|'

AFTER

Expr
    (AnonRecd
    (false, None,
     [(SynLongIdent ([a], [], [None]), Some (7,4--7,5),
       Const
         (Measure
            (Int32 1, (7,5--7,6),
             Seq ([Named ([m], (7,7--7,8))], (7,7--7,8)),
             { LessRange = (7,6--7,7)
               GreaterRange = (7,8--7,9) }), (7,5--7,9)))],
     (7,0--7,11), { OpeningBraceRange = (7,0--7,2) }), (7,0--7,11));

Checklist

• Test cases added
• Release notes entry updated

[github-actions]

❗ Release notes required

---

✅ Found changes and release notes in following paths:

Change path	Release notes path	Description
src/Compiler	docs/release-notes/.FSharp.Compiler.Service/10.0.100.md
LanguageFeatures.fsi	docs/release-notes/.Language/preview.md

[edgarfgp]

Fixed again another merge conflict :(

[edgarfgp]

@T-Gro This is ready

[T-Gro]

Thanks for finishing this @edgarfgp .
I do realize that my wishes to guard even simple improvements can be annoying at times, but our focus should be at the user who does not watch language evolution closely.

The situation with up-to-date local and outdated CI is quite real.When it comes to parser/lexer improvememnts, those errors (not reproducible locally) would be confusing (well, as they are currently before this change...) and hard to fix blindly.