Fix android module JAVA_NAMESPACE and JNI_NAMESPACE with unicode

[KeyWeeUsr]

If not switched to unicode, anything with formating such as + will trigger an error for py3 when accessing JAVA_NAMESPACE and likes:

b'1' + u'2'
u'1' + b'1'

and py2 just doesn't care.

[AndreMiras]

Traceback (most recent call last):
  File "/usr/lib/python3.6/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/usr/lib/python3.6/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/home/andre/workspace/EtherollApp/.buildozer/android/platform/python-for-android-new-toolchain/pythonforandroid/toolchain.py", line 1032, in <module>
    main()
  File "/home/andre/workspace/EtherollApp/.buildozer/android/platform/python-for-android-new-toolchain/pythonforandroid/toolchain.py", line 1028, in main
    ToolchainCL()
  File "/home/andre/workspace/EtherollApp/.buildozer/android/platform/python-for-android-new-toolchain/pythonforandroid/toolchain.py", line 550, in __init__
    getattr(self, args.subparser_name.replace('-', '_'))(args)
  File "/home/andre/workspace/EtherollApp/.buildozer/android/platform/python-for-android-new-toolchain/pythonforandroid/toolchain.py", line 149, in wrapper_func
    build_dist_from_args(ctx, dist, args)
  File "/home/andre/workspace/EtherollApp/.buildozer/android/platform/python-for-android-new-toolchain/pythonforandroid/toolchain.py", line 188, in build_dist_from_args
    build_recipes(build_order, python_modules, ctx)
  File "/home/andre/workspace/EtherollApp/.buildozer/android/platform/python-for-android-new-toolchain/pythonforandroid/build.py", line 635, in build_recipes
    recipe.prebuild_arch(arch)
  File "/home/andre/workspace/EtherollApp/.buildozer/android/platform/python-for-android-new-toolchain/pythonforandroid/recipes/android/__init__.py", line 33, in prebuild_arch
    bootstrap = bootstrap_name = self.ctx.bootstrap.name.decode('utf-8')
AttributeError: 'str' object has no attribute 'decode'

😢

[KeyWeeUsr]

I guess this is due to the Dockerfile using Python 2, therefore it worked for me on the device with py3 but it fails when you want to compile with py3 on the host. I think we should fully switch to unicode literals otherwise in case something is generated like variable = 'value' it'll always split to bytes on py2 and unicode on py3 and we either start checking each damn occurrence of this schrodinger's string or do this:

>>> b'x'.encode('utf-8').decode('utf-8')  # py2 + py3 bytes only
u'x'
>>> u'x'.encode('utf-8').decode('utf-8')  # py3 only
u'x'

and get kicked by anything that is not ASCII encodable.

Perhaps a fix for this would be to remove the decode() and enforce self.ctx.bootstrap.name to be always unicode with an assert.

[AndreMiras]

I'm in to fully switch since it's one step further the migration to python3 and we can still make it compatible with python2 with e.g. unicode_literals from future.
Edit: link to some a bug report kivy/buildozer#756

[AndreMiras]

Fixed by #1512

[KeyWeeUsr]

I have a proper fix with future package, but could not test it because of gradle crash. Will do asap.