Support BitMap64 ranges that end at the maximum value

[Shaisolaris]

The 64-bit range methods take an exclusive end as uint64_t, so a range that includes the largest value (end 2**64) raises OverflowError, while the 32-bit BitMap handles its own maximum:

from pyroaring import BitMap64
BitMap64().add_range(2**64 - 3, 2**64)          # OverflowError
BitMap64(range(2**64 - 3, 2**64))               # OverflowError

This affects add_range, remove_range, flip, flip_inplace, contains_range, range_cardinality and the range constructor path.

The change calls the closed roaring64_*_closed functions with end - 1, after checking that the end is between 0 and 2**64 and handling empty ranges first. Ends above 2**64, negative ends and non-integers still raise as before.

Tests: added TestRangeToMaximum in test.py. Locally (Python 3.12, Linux x86_64), with the new tests on the current master the 64-bit run has 4 of them failing; with this change they pass. test.py and test_state_machine.py otherwise give the same results before and after for both ROARING_BITSIZE=32 and 64. TestFFI::test_ffi_cardinality fails in my local build both before and after (an undefined symbol via ctypes), so I don't think it's related. I did not run other platforms.