Modernize gem for Ruby 3.0+

[eglitobias]

Bumps required_ruby_version to >= 3.0 and removes dead code and outdated idioms that accumulated since the 2.x days.

• Drop required_ruby_version to >= 3.0, update gemspec homepage to HTTPS, clean up Gemfile platform specifiers
• Remove RUBY_VERSION >= 1.9 guard and unused @range variable
• Modernize Ruby idioms: is_a? over ===, .nil? over == nil, key? over has_key?, raise X, msg over raise X.new(msg)
• Improve spec coverage for AMQP table type dispatch and frame handling (+14 examples, 90% → 91.9%)

[michaelklishin]

Thank you, @eglitobias!

[eglitobias]

@michaelklishin Thanks for merging! Just noticed the milestone is set to 2.7.0, but that version was already released on RubyGems with my previous bugfix. It's not showing up under GitHub Releases though.

[michaelklishin]

You are right, the latest release on the main page was shown as 2.6.0 because I haven't published 2.7.0.

[michaelklishin]

@eglitobias I don't have a good explanation without a flame graph but this results in a low double digit % improvement on the benchmarks. Most likely it's the x == nil and x.nil? change but I'll take it anyway.

[eglitobias]

nil? is a specialized nil check, while == nil uses normal == dispatch. In isolation that’s a micro-optimization, but it matters here because decode_header (frame.rb:65) runs on every received frame. That’s why a small per-call improvement shows up as a double-digit gain in the end-to-end benchmarks.

I did a quick local check with benchmark/ips:

require "benchmark/ips"

x = "hello"

Benchmark.ips do |b|
  b.report("== nil") { x == nil }
  b.report(".nil?")  { x.nil? }
  b.compare!
end

Ruby	== nil	.nil?	Speedup
3.4.5	17.2M i/s	30.7M i/s	1.78×
3.4.8	21.6M i/s	47.5M i/s	2.20×
4.0.2	21.5M i/s	46.1M i/s	2.14×

I’ll open a small follow-up PR to improve the benchmarking tooling and make it easier to compare results across Ruby versions.