improve param_value_size consistency

[bashbaug]

This fixes most of #1216.

• Improves the spec consistency for all of the param_value_size argument descriptions.
• Improves the spec consistency for all of the error conditions related to param_value_size.
• Fixes a few other minor editorial bugs found while making these changes.

[alycm]

LGTM, the param value changes is consistent with #1216 (and fundamentally makes sense).

The other misc error code clean ups all look good and non-controversial.

[alycm]

Suggested change

	* {CL_INVALID_VALUE} if an property name other than {CL_GL_CONTEXT_KHR},
	* {CL_INVALID_VALUE} if a property name other than {CL_GL_CONTEXT_KHR},

[bashbaug]

I've fixed this, though the fix was slightly different than the suggested change.

As background, the previous text said:

if an property name other than those specified in table 4.5 is specified in properties.

However, there is no table 4.5 in the current specs.

Going back to the old OpenCL 1.2 extensions spec, there was a table describing additions to table 4.5, which included the new properites to pass the CL_CGL_SHAREGROUP_KHR, CL_EGL_DISPLAY_KHR, etc., but these were additions to an existing table, not the full table.

Since the platform needs to be passed for ICD implementations at the very least to know where to dispatch this function, clearly the list of valid properties extends beyond just the additions to the table, but it's not immediately clear to me if all of the valid context properties from clCreateContext are valid here also, or if it's just some subset.

For now, I've simply changed the text to:

if a property name specified in properties is invalid.

This isn't great, but it's better than the previous text, which referred to a nonexistant table.