Align tensorArena in IMU_Classifier to 16 bytes

[Rainycat]

The TensorFlowLite library checks that the supplied arena memory is aligned to 16 bytes, and crashes if it is not. This change ensures the memory is aligned as expected.

There's a pull request already waiting, but it seems to be a more comprehensive (and more complicated!) fix. This might be a more simple way to get things working for now.

[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.}

[manchoz]

Hi @aentinger, both this PR and the fix suggested in #21 are mandatory to get the examples working properly in all the possibile scenarios.

I'm asking the user of #21 to create a PR to submit his fix.

Thanks!

[aentinger]

Hi @manchoz :wave: - sure - anything special you need me to do? (It appears you want me to merge this PR which I can, once the CLA is signed).

[manchoz]

Hi @Rainycat would you mind to sign the CLA? This is mandatory for approving the PR.

Thanks!

[Rainycat]

Hi @Rainycat would you mind to sign the CLA? This is mandatory for approving the PR.

Thanks!

Apologies for the delay @manchoz, hopefully all done now!

[aentinger]

This fixes #21.