Fix problem when reading terminal capabilities

An update to the terminfo entry for xterm-256color made between
ncurses-6.0+20170128-1 and ncurses-6.0+20170429-1 on Arch highlights a
problem with how terminal capabilities were being read and affects all
users of rb-readline on Arch.

The particular terminfo change was to alter the cr (carriage return)
capability from:

  ^M

to:

  \r

The rb-readline code that parses the capabilities was mistakenly storing
this capability as:

  \\r

causing "\r" to be displayed when entering a line of input into an
rb-readline program.

While I am reasonably confident that this change is sound, I am
concerned that there are other problems around the way terminal
capabilities are parsed because this patch fixes only this specific
case. Unfortunately, the `infocmp` line is a maze of regexes and escape
characters and checking and reworking it requires more time than I have
available just now.

Fixes #138.

Tested-by: Mark Somerville <mark@scottishclimbs.com> [Arch, MRI 2.4.1]
Signed-off-by: Mark Somerville <mark@scottishclimbs.com>

lib/rbreadline.rb

@@ -1812,7 +1812,7 @@ module RbReadline
   def get_term_capabilities(buffer)
     hash = {}
-      `infocmp -C`.split(':').select{|x| x =~ /(.*)=(.*)/ and hash[$1]=$2.gsub('\\E',"\e").gsub(/\^(.)/){($1[0].ord ^ ((?a..?z).include?($1[0]) ? 0x60 : 0x40)).chr}}
+      `infocmp -C`.split(':').select{|x| x =~ /(.*)=(.*)/ and hash[$1]=$2.gsub("\\r", "\r").gsub('\\E',"\e").gsub(/\^(.)/){($1[0].ord ^ ((?a..?z).include?($1[0]) ? 0x60 : 0x40)).chr}}
       @_rl_term_at7          =     hash["@7"]
       @_rl_term_DC           =     hash["DC"]
       @_rl_term_IC           =     hash["IC"]