Improve Element#attribute implementation as 6500x faster

[makenowjust]

Element#namespaces is heavy method because this method needs to traverse all ancestors of the element. Element#attribute calls namespaces redundantly, so it is much slower.

This PR reduces namespaces calls in Element#attribute. Also, this PR removes a redundant respond_to? because namespaces must return Hash in the current implementation.

Below is the result of a benchmark for this on my laptop.

RUBYLIB= BUNDLER_ORIG_RUBYLIB= /Users/makenowjust/Projects/github.com/makenowjust/simple-dotfiles/.asdf/installs/ruby/3.3.2/bin/ruby -v -S benchmark-driver /Users/makenowjust/Projects/github.com/ruby/rexml/benchmark/attribute.yaml
ruby 3.3.2 (2024-05-30 revision e5a195edf6) [arm64-darwin23]
Calculating -------------------------------------
                     rexml 3.2.6      master  3.2.6(YJIT)  master(YJIT)
   attribute_with_ns     425.420     849.271       5.336k       10.629k i/s -      1.000k times in 2.350620s 1.177481s 0.187416s 0.094084s
attribute_without_ns     834.750      5.587M      10.656k        2.950M i/s -      1.000k times in 1.197963s 0.000179s 0.093846s 0.000339s

Comparison:
                attribute_with_ns
        master(YJIT):     10628.8 i/s
         3.2.6(YJIT):      5335.7 i/s - 1.99x  slower
              master:       849.3 i/s - 12.52x  slower
         rexml 3.2.6:       425.4 i/s - 24.98x  slower

             attribute_without_ns
              master:   5586593.2 i/s
        master(YJIT):   2949854.4 i/s - 1.89x  slower
         3.2.6(YJIT):     10655.8 i/s - 524.28x  slower
         rexml 3.2.6:       834.8 i/s - 6692.53x  slower

This result shows that Element#attribute is now 6500x faster than the old implementation if namespace is not supplied.

It seems strange that it is slower when YJIT is enabled, but we believe this is a separate issue.

Thank you.

[kou]

attribute_ is redundant because this file is attribute.yaml.
Can we use with_ns/without_ns or something?

[kou]

Suggested change

	prefix = namespaces.key(namespace) if namespace
	prefix = namespaces[namespace] if namespace

may be faster than namespaces.key(namespace).

[makenowjust]

namespaces.key(namespace) means namespaces.invert[namespace]. Your suggestion changes the code's meaning.

[kou]

Ah, sorry.