This is a simple matter of transforming a time in seconds into a corresponding sample count.
You can easily compute the duration in seconds for one bar at your given BPM. Then, the sample rate as reported to you in the prepare callback says you how many samples represent one second of audio. Multiplying that rate with the desired interval in seconds gives you the exact sample count you need. Pre-Allocate an audio buffer of that size in your prepare method and copy over the desired number of samples in the processBlock callbacks from the buffer passed to you. It might be possible that it would take multiple proccessBlock callbacks or even one processBlock callback could contain more data then you need, you have to handle all scenarios.
By the way, the BitRate is irrelevant for you, this is only about the dynamic resolution, not the time resolution of your audio. Furthermore, inside a plugin, audio is alway processed as normalised 32Bit floating point values (64Bit doubles are possible too, but not the standard), so from that point of view, the BitRate is always fixed