you need to provide a picture - also its a NL analysis and failure is tricky to pull off.
I have tried shear-failure pressure-relief devices.
Bill is correct that, unless as per Mike's assumption of very small shear deflection, there will be a shift of the load from pure shear to partial tensile load and that will throw off the assumption of failure solely to a uniaxial shear, requiring nonlinear analysis for improved results.
Since you are not wanting to know the pressure at which shear starts, but rather the point at which you see rupture, your stress level at rupture will be more closely related to UTS than yield; so if you use Mike's methodology you'll need to jigger your material properties to reflect that. As well, you'd best use the UTS from the mill test report for your actual cylinder head material (because typical values are statistical and standard minimums are frequently exceeded by large margins) Even then, you're likely to have a fair amount of uncertainty. If you really need an accurate knowledge of the failure point you may need to make a few heads and do some destructive tests to calibrate your materials to your geometry or to use pre-engineered solutions. The last takes all the fun out of engineering this component but might be the way to go.
I mean physically cut the part--either add a configuration or save out a copy for analysis. You can insert symmetry as a boundary condition, or you can constrain both symmetry surfaces so that they can't move in the normal direction. Either way will have the same effect.