Add casting and relabeling to SeqType coercion

[GuntherRademacher]

These changes extend SeqType.coerce by the new rules for coercing to xs:anyURI, xs:base64Binary and xs:hexBinary, as well as for relabeling.

This fixes these QT4 test cases:

• DynamicFunctionCall-080
• DynamicFunctionCall-087
• DynamicFunctionCall-130
• DynamicFunctionCall-131
• DynamicFunctionCall-132
• DynamicFunctionCall-137
• FunctionCall-400
• FunctionCall-405
• function-call-promotion-401
• function-call-promotion-402
• function-call-promotion-404
• function-call-promotion-406
• function-call-promotion-407
• function-call-promotion-408
• function-call-promotion-409
• K2-FunctionProlog-5a
• K2-FunctionProlog-6a

Test case DynamicFunctionCall-130 also failed due to JDK-8272702, so a workaround for this was added to Uri.resolve.

[GuntherRademacher]

The second commit adds coercion from xs:double to xs:float, and from xs:double and xs:float to xs:decimal. This additionally fixes these test cases:

• xs-double-403
• xs-double-404
• xs-double-405
• xs-double-406
• DynamicFunctionCall-126
• DynamicFunctionCall-127

hof-919 now fails, but this is due to expecting a coercion failure, which no longer occurs with the new rules. I have raised a QT4 issue about this (qt4cg/qt4tests#78).

[ChristianGruen]

I’ve started to unify the coercion rules, based on this PR.

[ChristianGruen]

Merged (382371274736629d2722fdd083513e7f6692d264, 280970b5e2e2e03c71e1cefa7dc1737aa91b5f21, 896d315fc8a13a69436c9cc6f52dd4c9a7748dd9)