Support copy command on windows and wsl

[hogelog]

Enable the copy command on Windows and WSL environments.

The copy command now works both on Windows Ruby and on WSL, copying the evaluated result to the system clipboard.

Examples

Windows (mingw-ucrt)

PS C:\Ruby\repos\irb> ruby -v
ruby 3.4.8 (2025-12-17 revision 995b59f666) +PRISM [x64-mingw-ucrt]
PS C:\Ruby\repos\irb> ruby -Ilib .\exe\irb
irb(main):001> RUBY_PLATFORM
=> "x64-mingw-ucrt"
irb(main):002> copy 2**11
Copied to system clipboard
irb(main):003> 2048 # <- Ctrl-V
=> 2048

WSL

hogelog@lenovo:/mnt/c/Ruby/repos/irb$ ruby -Ilib ./exe/irb
irb(main):001> RUBY_PLATFORM
=> "x86_64-linux"
irb(main):002> copy 2**10
Copied to system clipboard
irb(main):003> 1024 # <- Ctrl-V
=> 1024

Windows workaround

On Windows, passing text via stdin does not work reliably in IRB. After calling IO.popen, IRB’s input handling becomes unstable, but I have not fully identified the root cause yet.

To work around this, the Windows implementation invokes PowerShell’s Set-Clipboard and passes the text via command-line arguments instead of stdin:

powershell.exe -NoProfile -Command Set-Clipboard -Value <text>

This approach works reliably on native Windows Ruby.

This issue is likely not limited to the copy command, and there may be other cases where IRB does not behave correctly. If this problem can be resolved, it should be possible to simply call clip.exe even on Windows. I plan to investigate this separately.

[tompng]

There's a flag --disable-gems

$ ruby --disable-gems -e binding.irb      
irb(main):001> Gem
uninitialized constant Gem (NameError)

I think it's better to add a guard like defined?(Gem) &&

[hogelog]

I had not considered running IRB in an environment where Gem is not loaded. But I think a simple Windows check is sufficient here. How about this diff?
d97375e

[tompng]

On Windows, passing text via stdin does not work reliably in IRB

This might be fixed in ruby/reline#875
Although I also think adding this workaround for a while is worth. What do you think?

[hogelog]

I had not noticed ruby/reline#875. With this fix, the workaround seems unnecessary, and it looks like we can remove the unnatural branching, so I deleted it e90574a