1) It is a litteral, where it is a % then a r(regular expression), w(array), q(string) etc to denote different litterals. 
2)
ruby-1.9.2-p136 :001 > %w{1 2 3}
 => ["1", "2", "3"] 
ruby-1.9.2-p136 :002 > %w[1 2 3]
 => ["1", "2", "3"] 
ruby-1.9.2-p136 :008 > %w!a s d f!
 => ["a", "s", "d", "f"] 
ruby-1.9.2-p136 :009 > %w@a s d f@
 => ["a", "s", "d", "f"] 
So you can see that you can use any charater as long as it marks both the beginning and end of the content.
3)
Here are some other examples:
Strings:(%q or %Q)
ruby-1.9.2-p136 :016 > %Q[ruby is cool]
 => "ruby is cool" 
ruby-1.9.2-p136 :017 > %q[ruby is "cool"]
 => "ruby is \"cool\"" 
Regex: (%r)
ruby-1.9.2-p136 :019 > %r[(\w+)]
 => /(\w+)/ 
Sys command: (%x)
ruby-1.9.2-p136 :020 > %x[date]
 => "Tue Mar 29 12:55:30 EDT 2011\n"
4) They cannot be nested because the %w means white space divided array. So if you try to do multi level, it would look like this:
ruby-1.9.2-p136 :003 > %w{1 %w{2 3 4} 5}
 => ["1", "%w{2", "3", "4}", "5"] 
To accomplish this, you would need to use the more verbose syntax:
ruby-1.9.2-p136 :011 > [1, [2,3,4], 5]
 => [1, [2, 3, 4], 5]