Revert "test-bot: revise Step output transcoding and XML character filtering"

This reverts commit ceed6636d6986bb2b993ba4680ae484deb76ae29.

```
==> git clean -ffdx
HEAD is now at ceed663 test-bot: revise Step output transcoding and XML character filtering
Removing Library/Homebrew/test/.bundle/
Removing Library/Homebrew/test/fs_leak_log
Removing Library/Homebrew/test/vendor/
Error: invalid regular expression: /[^

 -uD7FFuE000-uFFFDu{10000}-u{10FFFF}]/
```
Using Ruby 1.8.

Library/Homebrew/cmd/test-bot.rb

@@ -9,17 +9,16 @@
 # --skip-setup:    Don't check the local system is setup correctly.
 # --skip-homebrew: Don't check Homebrew's files and tests are all valid.
 # --junit:         Generate a JUnit XML test results file.
-# --no-bottle:     Run brew install without --build-bottle.
+# --no-bottle:     Run brew install without --build-bottle
 # --keep-old:      Run brew bottle --keep-old to build new bottles for a single platform.
-# --legacy         Build formula from legacy Homebrew/homebrew repo.
+# --legacy         Bulid formula from legacy Homebrew/homebrew repo.
 #                  (TODO remove it when it's not longer necessary)
 # --HEAD:          Run brew install with --HEAD
-# --local:         Ask Homebrew to write verbose logs under ./logs/ and set HOME to ./home/.
-# --tap=<tap>:     Use the git repository of the given tap.
+# --local:         Ask Homebrew to write verbose logs under ./logs/ and set HOME to ./home/
+# --tap=<tap>:     Use the git repository of the given tap
 # --dry-run:       Just print commands, don't run them.
 # --fail-fast:     Immediately exit on a failing step.
-# --verbose:       Print test step output in realtime. Has the side effect of passing output
-#                  as raw bytes instead of re-encoding in UTF-8.
+# --verbose:       Print out all logs in realtime
 # --fast:          Don't install any packages but run e.g. audit anyway.
 #
 # --ci-master:           Shortcut for Homebrew master branch CI options.
@@ -39,10 +38,6 @@ module Homebrew
   BYTES_IN_1_MEGABYTE = 1024*1024
   HOMEBREW_TAP_REGEX = %r{^([\w-]+)/homebrew-([\w-]+)$}
 
-  def ruby_has_encoding?
-    String.method_defined?(:force_encoding)
-  end
-
   def resolve_test_tap
     if tap = ARGV.value("tap")
       return Tap.fetch(tap)
@@ -147,9 +142,7 @@ module Homebrew
       end
 
       verbose = ARGV.verbose?
-      # Step may produce arbitrary output and we read it bytewise, so must
-      # buffer it as binary and convert to UTF-8 once complete
-      output = Homebrew.ruby_has_encoding? ? "".encode!("BINARY") : ""
+      @output = ""
       working_dir = Pathname.new(@command.first == "git" ? @repository : Dir.pwd)
       read, write = IO.pipe
 
@@ -162,14 +155,13 @@ module Homebrew
           working_dir.cd { exec(*@command) }
         end
         write.close
-        while buf = read.readpartial(4096)
+        while buf = read.read(1)
           if verbose
             print buf
             $stdout.flush
           end
-          output << buf
+          @output << buf
         end
-      rescue EOFError
       ensure
         read.close
       end
@@ -179,9 +171,8 @@ module Homebrew
       @status = $?.success? ? :passed : :failed
       puts_result
 
-
-      unless output.empty?
-        @output = fix_encoding(output)
+      if has_output?
+        @output = fix_encoding(@output)
         puts @output if (failed? || @puts_output_on_success) && !verbose
         File.write(log_file_path, @output) if ARGV.include? "--keep-logs"
       end
@@ -191,11 +182,11 @@ module Homebrew
 
     private
 
-    if Homebrew.ruby_has_encoding?
+    if String.method_defined?(:force_encoding)
       def fix_encoding(str)
+        return str if str.valid_encoding?
         # Assume we are starting from a "mostly" UTF-8 string
         str.force_encoding(Encoding::UTF_8)
-        return str if str.valid_encoding?
         str.encode!(Encoding::UTF_16, :invalid => :replace)
         str.encode!(Encoding::UTF_8)
       end
@@ -953,13 +944,11 @@ module Homebrew
 
           if step.has_output?
             # Remove invalid XML CData characters from step output.
-            valid_xml_chars = "\x09\x0A\x0D\x20-\uD7FF\uE000-\uFFFD\u{10000}-\u{10FFFF}"
-            output = step.output.gsub(/[^#{valid_xml_chars}]/, "\uFFFD")
+            output = step.output.delete("\000\a\b\e\f\x2\x1f")
 
             if output.bytesize > BYTES_IN_1_MEGABYTE
-              slice_start = [0, -BYTES_IN_1_MEGABYTE].max
-              output = "truncated output to ~1MB:\n" \
-                + output.slice(slice_start, output.length - slice_start)
+              output = "truncated output to 1MB:\n" \
+                + output.slice(-BYTES_IN_1_MEGABYTE, BYTES_IN_1_MEGABYTE)
             end
 
             cdata = REXML::CData.new output