Backend.ai

------------------

- Update the default image metadata/aliases to include latest deep learning kernels.

------------------

- Fix creation of GPU sessions with GPU resource limits unspecified in the
client-side. The problem was due to a combination of misconfiguration
(image-metadata.yml) and mishandling of "None" values with valid dictionary keys.
- Update coding style rules and the flake8 package.

------------------

- Fix a critical regression bug of tracking available memory (RAM) of agents due to
changes to relative resource shares from absolute resource amounts.
- Backport a temporary patch to limit the maximum number of kernel execution records
returned by the admin GraphQL API (until we have a proper pagination support).
- Update the list of our public kernel images as we add support for latest TensorFlow
versions including v1.10 and v1.11 series. More to come!

------------------

- Support CORS (cross-origin resource sharing) for browser-based API clients (99).
- Fix the agent revival detection routine to update agent's address and region
for movable demo devices (100).
- Update use of deprecate APIs in our dependencies such as aiohttp and aiodocker.
- Let the config server to refresh configuration values from etcd once a minute.

------------------

- Generalizes accelerator supports
- Accelerators such as CUDA GPUs can be installed as a separate plugin (66)
- Adds support for nvidia-docker v2 (64)
- Adds support for allocation of multiple accelerators for one kernel container as
well as partial shares of each accelerator (66)
- Revamp the agent restart and kernel initialization processes (35, 73)
- The view of the agent can be limited to specific CPU cores and GPUs
using extra CLI arguments: `--limit-cpus`, `--limit-gpus` for
debugging and performance benchmarks. (65)

-------------------

- Add CORS support (Hotfix 99 backported from v1.4 and master)