read_depth_offset: new test to check depth offset calculation

We have tests to validate depth offsets in general, but these apply to the default depth buffer (usually 24 bit fixed point). This test checks the details of the depth offset calculation by using NV_read_pixel to read back the written values, and verifies that e.g. the minimum resolvable difference used for floating point buffers matches the spec. This revealed a bug on panfrost.

Signed-off-by: Eric R. Smith eric.smith@collabora.com