Enhance Clover database to keep full instrumentation history

Problem:

Currently Clover database keeps list of instrumentation sessions plus information about classes which is a sum of all previous instrumentation sessions in such way that for given class only the latest class version is stored.

There are two limitations of current design:

1) Test optimization might not work correctly for some test frameworks for Groovy language. This is due to nature of the Groovy language itself, where compiler allows to compile code which does not have all symbols resolved (they're being resolved at runtime). As a consequence if a dependent application class is changed, this does not to have trigger recompilation of the test class. See comment for more details.

2) Syntax highlighting in HTML report is always based on the latest class version, so if someone has older source file, it can be rendered incorrectly. There's a warning for this in HTML report, but we could fetch the actual line/char indexes from the older class version.

Solution:

1) Enhance UpdateableRegFile and keep all versions of classes. This should be optional and disabled by default.

2) Enhance RegFile API to allow querying for older versions of given file (by default it should always return the latest version).

3) Add new option for <clover-setup> and <clover-instr> to allow to keep full instrumentation history. Either as "true/false" or as a number (maximum number of versions of the single class to store).

4) Add API to calculate index bitset as a logical OR of bitsets from all versions of the given file or class.