why acquire/release are used several times in the init function ?

I looked at the code more in depth, and this is perfectly fine in term of workflow: it's just that some private functions are not acquiring the bus before performing their read/write operations.

here is an output with a nucleo-l073:

2018-06-21 11:25:57,327 - INFO # Connect to serial port /dev/ttyACM0
Welcome to pyterm!
Type '/exit' to exit.
2018-06-21 11:25:58,801 - INFO # H: 56.4%, T: 24.8°C
2018-06-21 11:26:00,805 - INFO # H: 56.1%, T: 24.8°C
2018-06-21 11:26:02,803 - INFO # H: 56.0%, T: 24.8°C
2018-06-21 11:26:04,807 - INFO # H: 55.9%, T: 24.8°C

I looked at the code more in depth, and this is perfectly fine in term of workflow: it's just that some private functions are not acquiring the bus before performing their read/write operations.

Sure, but this is first time I saw this. Is there any gain from doing this ? I don't have strong opinion about this. If the driver was already in this state before, this is not our rights to change it with the I2C refactoring.
Anyway, it would be great if someone could test it, otherwise I can leave an untested ACK.

it would be great if someone could test it

I think @smlng has the hardware, since he initially wrote this driver. Maybe he can comment on your other questions.