[DO NOT MERGE] Fix Tensorflow Lite GPU inference

[koleini]

No description provided.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[filipecosta90]

@koleini worth noticing that on the tests we've discussed with mobilenet, TFLITE v2.3.2, and GPU backend, using this branch with the latest master changes merged, for tensors that are a large batch of images ( >=16 ) we experience crashes on RedisAI:

6705:M 21 Mar 2021 16:11:39.497 * <ai> TFLITE backend loaded from bin/linux-arm64v8-release/install-gpu/backends/redisai_tflite/redisai_tflite.so


=== REDIS BUG REPORT START: Cut & paste starting from here ===
6705:M 21 Mar 2021 16:11:39.662 # Redis 6.2.1 crashed by signal: 11, si_code: 2
6705:M 21 Mar 2021 16:11:39.662 # Accessing address: 0x7f94744000
6705:M 21 Mar 2021 16:11:39.662 # Killed by PID: -1804320768, UID: 127
6705:M 21 Mar 2021 16:11:39.662 # Crashed running the instruction at: 0x7f9ac582b8

------ STACK TRACE ------
EIP:
/lib/aarch64-linux-gnu/libc.so.6(+0x832b8)[0x7f9ac582b8]

Backtrace:
linux-vdso.so.1(__kernel_rt_sigreturn+0x0)[0x7f9ae796c0]
/lib/aarch64-linux-gnu/libc.so.6(+0x832b8)[0x7f9ac582b8]
bin/linux-arm64v8-release/install-gpu/backends/redisai_tflite/redisai_tflite.so(tfliteRunModel+0x458)[0x7f9ae3a588]
bin/linux-arm64v8-release/install-gpu/backends/redisai_tflite/redisai_tflite.so(RAI_ModelRunTFLite+0x188)[0x7f9ae35520]