That's calculable by fixing your conversion or picture style parameters and varying only exposure of a fixed scene, but it is going to vary through the histogram with most histograms, which use gamma or logarithms on the right side, and are more linear on the left. You couldn't even have black or zero on the left end if the left side was logarithmic, as there is no such thing as the logarithm of zero; a pure logarithmic histogram would go on forever to the left. I would assume that in cameras, the histogram is almost always directly based on an 8-bit converted image, and the values in that image (or a binning or subsampling of it). With such histograms, the number of histogram pixels per stop should peak in the middle of the range, and always be lower on the left end, and lower on the right end when the converter rolls in and compresses an extended range of highlights.
Here's a crude, quantized guess-timation of how objects exactly in a 1-stop luminance series might render with a highlight-extending picture style; just keep in mind that the very left section would actually be denser with the "|" lines, if you kept halving exposure:
|||| | | | | | | | | ||