Add CUDA 10.2 support

[lazystone]

Description

I tried to run this demo to get familiar with djl, but could not run training on GPU because I have CUDA 10.2 installed:

Caused by: java.lang.UnsatisfiedLinkError: Unable to load library '/home/andrej/.mxnet/cache/1.6.0-a-20191127cu101mkl-linux-x86_64/libmxnet.so':
libcudart.so.10.1: cannot open shared object file: No such file or directory
libcudart.so.10.1: cannot open shared object file: No such file or directory
Native library (home/andrej/.mxnet/cache/1.6.0-a-20191127cu101mkl-linux-x86_64/libmxnet.so) not found in resource path (/home/andrej/projects/djl-demo/build/classes/java/main:/home/andrej/projects/djl-demo/build/resources/main:/home/andrej/.gradle/caches/modules-2/files-2.1/ai.djl.mxnet/mxnet-model-zoo/0.2.0/355dfb3163430430f25c7e86267fc5c621276179/mxnet-model-zoo-0.2.0.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/ai.djl.mxnet/mxnet-engine/0.2.0/7841bd1c3fc2f44fe76cb2e8a083dfead4de3f9a/mxnet-engine-0.2.0.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/net.java.dev.jna/jna/5.3.0/4654d1da02e4173ba7b64f7166378847db55448a/jna-5.3.0.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/org.apache.httpcomponents/httpcore/4.4.12/21ebaf6d532bc350ba95bd81938fa5f0e511c132/httpcore-4.4.12.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/commons-cli/commons-cli/1.4/c51c00206bb913cd8612b24abd9fa98ae89719b1/commons-cli-1.4.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/org.slf4j/slf4j-simple/1.7.29/82ae07f95088577987a15d90171de12b00d81847/slf4j-simple-1.7.29.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/org.apache.commons/commons-csv/1.7/cb5d05520f8fe1b409aaf29962e47dc5764f8f39/commons-csv-1.7.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/ai.djl/basicdataset/0.2.0/fa73e42fb774b56f23a030a6c95159a1987d8110/basicdataset-0.2.0.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/ai.djl/model-zoo/0.2.0/dbe300ddc19ec809002ed9a6214dac11e39a1055/model-zoo-0.2.0.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/ai.djl/repository/0.2.0/266c3a327e89b82234c03a713f05067567c2e9dd/repository-0.2.0.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/com.google.code.gson/gson/2.8.5/f645ed69d595b24d4cf8b3fbb64cc505bede8829/gson-2.8.5.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/ai.djl/api/0.2.0/c83672c1e7178830ea9c43b98603d5fa7737fd78/api-0.2.0.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/ai.djl.mxnet/mxnet-native-cu101mkl/1.6.0-a/c67432f4f6ba4273a13c3f9efff52e5f2710c888/mxnet-native-cu101mkl-1.6.0-a-linux-x86_64.jar:/home/andrej/.gradle/caches/modules-2/files-2.1/org.slf4j/slf4j-api/1.7.29/e56bf4473a4c6b71c7dd397a833dce86d1993d9d/slf4j-api-1.7.29.jar)
        at com.sun.jna.NativeLibrary.loadLibrary(NativeLibrary.java:302)
        at com.sun.jna.NativeLibrary.getInstance(NativeLibrary.java:455)
        at com.sun.jna.Library$Handler.<init>(Library.java:192)
        at com.sun.jna.Native.load(Native.java:596)
        at com.sun.jna.Native.load(Native.java:570)
        at ai.djl.mxnet.jna.LibUtils.loadLibrary(LibUtils.java:80)
        at ai.djl.mxnet.jna.JnaUtils.<clinit>(JnaUtils.java:68)
        at ai.djl.mxnet.engine.MxEngine.<init>(MxEngine.java:36)
        at ai.djl.mxnet.engine.MxEngineProvider.<clinit>(MxEngineProvider.java:21)
        at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
        at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62)
        at java.base/jdk.internal.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
        at java.base/java.lang.reflect.Constructor.newInstance(Constructor.java:490)
        at java.base/java.util.ServiceLoader$ProviderImpl.newInstance(ServiceLoader.java:779)
        ... 9 more
        Suppressed: java.lang.UnsatisfiedLinkError: libcudart.so.10.1: cannot open shared object file: No such file or directory
                at com.sun.jna.Native.open(Native Method)
                at com.sun.jna.NativeLibrary.loadLibrary(NativeLibrary.java:191)
                ... 22 more

CUDA 10.2 seems to be mainstream at the moment and it's easier to find installation instructions for this version.

Rolling back to CUDA 10.1 might be troublesome and will repel some users from using this library.

Will this change the current api? How?

No

Who will benefit from this feature?

All potential users of this djl including myself

[frankfliu]

@lazystone Thanks for reaching to us. We are working on 10.2 support. Will add 10.2 support soon.

For the time being, you can build mxnet with cuda102 from source, and export MXNET_LIBRARY_PATH to point to your libmxnet.so file. DJL can pick up libmxnet.so from there.

[lazystone]

I'll wait until CUDA 10.2 works out of the box :)

[frankfliu]

@lazystone
Now DJL support CUDA 10.2 in master branch. 0.3.0 release will officially support CUDA 10.2