…r#323

* Make the default rehash impl no-op. If the map supports rehash,
  the correct Java method will be bound instead.
* Make the default shift have a better error explaining that Java
  Map does not generally preserve insertion order.

Allows using respond_to? to check if they do anything. #3142

Added uri to path when using file protocol to chdir

* See #2867.

- added tests for next month with calendar reform
- uses solution provided in #1769 comments
- also fixes issues #1769

When the block passed to map makes modifications to the array
under iteration, we may prematurely finish the map loop due to the
size changing. However our logic for creating the mapped array
assumed the new array's size would always be the same as the
original array's size, leading to an array with null elements.
This fix uses the iteration count as the final size, so we at
least know how many elements in the new array were populated.

Note that this behavior is officially undefined; modifying the
array while performing internal iteration can cause peculiar
effects across runtimes and potentially across the different
versions of the same runtime. We add a regression spec here to
at least make sure we don't produce an invalid array.

... and handle proc's with arity >

than the functional-interface methods parameter count

avoids incorrect ambiguous warning e.g. in case of :
`java.io.File.new('.').listFiles { |pathname| ... }`

…ble/assignable found

this avoids an incorrect ambiguous warning when matching proc-to-iface methods e.g. : 
`executor_spec.rb:22 warning: ambiguous Java methods found, using submit(java.util.concurrent.Callable)`

... also exploded the 3 level nested for loop to be only 2 levels

… (thus not matching)

…ilure at travis-ci)

…on getMethods order

…ases correctly

... they were still dependent on getMethods order (travis-ci failures in **spec:ji**)

…vaMethod

…Closes #3163

MRI checks if the EXCL flag is Qnil to see if the range has been initalized. This doesn't work
in Java because we use a boolean which can't have a nil value. Instead, introduce a new isInited
flag which is used to track the initialization state of the Range, which corrently permits #dup
while not permitting reinitialization of a Range.

* The currentThread ThreadLocal will tell us anyway.

…kill.

* Just use #to_s if you need to print an object before bootstrap/kernel.rb is loaded.

Fix RubyRange#initialize_copy mistakenly erroring on #dup'ing a range

* It is about 292 471 208 years.

* The thread can also have the status "sleep" very shortly when calling Mutex#lock.

* The thread can also have the status "sleep" very shortly when calling Mutex#lock.

It was giving bad rusults for cases like:
{a:1, b:2}.merge(b:3) # => {:a=>1, :b=>2}

* to be able to get the backtrace in debuger

…e instance

from core/rubinius/boostrap/thread to core/rubinius/api/shims/thread

…::Primitive