Some groups looking into eQTL (expression quantitative trait loci, where a SNP allele associates with increased or decreased levels of a mRNA) have used a distance of 1 million bp from SNP to gene in order to assign that association as cis.

The MYC enhancer, functional in human and mouse prostate tissue in an allele-specific manner, as described by Wasserman Nobrega et al (2010 Genome Res) is 340 kbp from the MYC TSS.

The OCA2 enhancer, shown as well to have allele-specific activity with regard to pigmentation in human (Visser Kayser et al 2012 Genome Res), maps within HERC2 and lies 21 kbp from the OCA2 TSS. (Of note, both MYC and OCA2 examples are funtional follow-ups to GWAS hits.)

Other responses could provide still more examples of course, but at least here you have a range of examples from the very distant to the rather near.

Good question. I am not aware of a reference on this, but would be interested to know if any else has one.

I would take genome annotation and taget gene information from the ORegAnno database and compute this yourself.