Depending on your peak caller, you may have metrics like total number of reads or peak summit height that you could summarize directly from the BED file of called peaks. These numbers are likely to be more accurate since they are usually calculated after shifting the reads by the average fragment size (see Fig 1 in the SPP paper if this is unclear) and may already be normalized to input and library size.

That said, 1-bp overlap seems to be the most commonly used for interval overlaps (in the absence of any biological arguments saying otherwise).

Honestly, I haven't done it for ChIP-seq but did it for WGS CNVs. There is a tool called CNV-seq which takes 2 columns of your SAM file, determines the window size and bins the reads across the genome. Even though it is for CNVs I think you can still use it to look at the average coverage as well.

There is also another software CHANCE which would give you an estimation of the percentage of the genome enriched for your signal if you feed it control and treatment BAM/SAM files.

Hope this helps!