#10819)

Split argv at the first `--` and only post-process the compiler-args prefix; the suffix (including the `--` separator itself) reaches downstream handlers byte-for-byte. Keeping `--` in the suffix preserves two paths:

* with a script preceding `--`, the script's OptionGeneral IsScript handler captures the suffix as script args (matching pre-fix behaviour for fsi.CommandLineArgs);

* with no script (e.g. `dotnet fsi -- -d 5`), the `--` token reaches ParseCompilerOptions and fires its OptionRest recordExplicitArg handler so `-d` and `5` are recorded as explicit args instead of being parsed as compiler options.

Updates the regression test theory to expect `--` in the captured tail and adds a new no-script Fact that locks the OptionRest path.

Co-authored-by: Copilot <[email protected]>

Fantomas-applied formatting in src/Compiler/Interactive/fsi.fs (whitespace
and pattern-match line collapse only; no behavioural change) and a single
new bullet under ### Fixed in docs/release-notes/.FSharp.Compiler.Service/11.0.100.md
referencing Issue #10819.

Validation: focused regression filters (FsiCommandLineArgsTests,
CommandLineArgs, FsiCliTests) green under -c Release on both net10.0 and
net48. SurfaceAreaTest passes on net10.0 with no baseline update (net48
has zero tests for that filter - platform-conditional). Full
FSharp.Compiler.ComponentTests and FSharp.Compiler.Service.Tests suites
were run under -c Release as the documented alternative to
'./build.sh -c Release --testcoreclr'; the only failures observed were
two well-known flaky ScriptOptionsTests nuget-resolution tests on net48
(unrelated to argument handling; confirmed flaky in isolation; net10.0
runs all 4364 tests green).

Co-authored-by: Copilot <[email protected]>