Correct core/stdlib signatures to match Ruby 4.0

[zonuexe]

Summary

Several core/stdlib signatures were inaccurate against the Ruby 4.0 implementation. This corrects them.

Ruby Method	Change
IO#pathconf	Add the method — (Integer name) -> Integer?
Digest::Instance#finish	Return String, not self
Digest::Class.digest / .hexdigest / .bubblebabble	Accept trailing *args (passed through to the constructor)
Digest::Class.file / .base64digest	Accept trailing *args
Etc.sysconf	Return type is nilable — -> Integer?
Etc.systmpdir	Return type is nilable — -> String?
Etc::Group.each / Etc::Passwd.each	Add block / Enumerator overloads instead of -> untyped
Etc::Passwd	Add the age, comment, and quota members (with their setters)
StringIO.open	Add the non-block overload — (?String, ?String?) -> StringIO
StringIO#fcntl	Always raises NotImplementedError — return bot
StringIO#fsync	Always returns 0 — -> Integer (not Integer?)
StringIO#internal_encoding	Always returns nil
StringIO#external_encoding	Return type is nilable — -> Encoding?
StringIO#pread	Add the method
StringIO#read_nonblock	Accept exception:; return type is nilable
StringIO#write_nonblock	Add the method
StringIO#sysread	outbuf is optional
StringIO#ungetc	Also accepts an Integer
StringIO#set_encoding_by_bom	Add the method

Notes

• Monitor#wait_for_cond / MonitorMixin::ConditionVariable#wait (-> bool) are intentionally excluded here — they are covered by Replace deterministic untyped return types with concrete types #2966.
• rdoc comments for the newly added methods match rbs annotate output (confirm_annotation).

Test plan

• bundle exec rbs validate
• bundle exec rbs -r digest -r etc -r stringio validate
• bundle exec rubocop core/io.rbs stdlib/digest/0/digest.rbs stdlib/etc/0/etc.rbs stdlib/stringio/0/stringio.rbs

[ksss]

I have already merged it, but I believe there is an issue with these changes.
For example, the following code will cause an ArgumentError in actual behavior, but it passes the type check.

Digest::SHA256.base64digest("aaa", :bbb) #=> ArgumentError

While these methods are indeed delegated to the constructor, since it is not possible to describe this delegation in RBS, I believe we should add appropriate overloads in each class.

I will fix it soon.

We will address the issue that the existing test cases failed to detect the problem as a problem with the test harness.

[ksss]

Done #2980