i2c:ocores: fixes and polling mechanism

[Federico Vaga]

The first two patches fix what I believe are bugs.

The third patch add a polling mechanism for those systems where interrupts
are not available.

All these patches have been tested on a system without interrupt, this
means that I used my third patch to validate also the other two.
I would be nice if someone can run verify this also on other system,
perhaps with interrupts. If you consider it a useful information, I'm not
using devicetree for this installation.