Update depends on to allow [OR|XOR|Array] or [Cask|Formula]

[BenjaminHCCarr]

With the changes to brew specifically with regard to the boneyard changes, this week I had a bunch of formulas pop up as deleted. Mostly due to the :tuntap signed kext problem:
Homebrew/homebrew-core/pull/2075
Homebrew/homebrew-core/pull/2075
An example: Homebrew/homebrew-core/commit/17df29ee0

We have a bunch of problems that could all be solved by changing/adding stanzas, re: depends_on and meta-formulas:
/issues/2300 [meta-formula]
@MPLew-is posited MPLew-is/homebrew-blend
he also noted

From

brew/Library/Homebrew/requirement.rb

Lines 34 to 43 in b6d1076

	def message
	_, _, class_name = self.class.to_s.rpartition "::"
	s = "#{class_name} unsatisfied!\n"
	if cask
	s += <<-EOS.undent
	You can install with Homebrew-Cask:
	brew cask install #{cask}
	EOS
	end

Homebrew/homebrew-cask/issues/19369 [OR / XOR]
Homebrew/homebrew-cask/issues/26345 [OR / XOR & Formula vs Cask (R)]
/issues/1887 [mariadb vs mysql; breaks doctor]

While we have discussed the use of caveats in caskroom/homebrew-cask as a "workaround" this is less than ideal; And if testing isn't possible as mentioned in Homebrew/homebrew-cask/issues/15603 and Homebrew/homebrew-core#2076 (comment)

Maybe we require some commitment by end user; I don't mean to create more work; but possibly a flag --untested for something that requires a kext like tuntap on a formula.

It seems we have hit a point where we have multiple problems that require similar fixes to the depends_on are; either allowing an array (like all mysql or redis forks) or relying on cask alone (tuntap), or relying on cask or formula (R).

[MikeMcQuaid]

We may consider allowing depending on Casks in the future but this isn't something we're actively planning to work on.