squashed commits:

create db->get_key_after_bytes() api down to the ft layer, and start a unit test
refs #18

setup/teardown for get_key_after_bytes test
refs #18

rename test because it really is going to be a small unit test
refs #18

implement a few initial checks
refs #18

adding actually_skipped parameter because we will want it in splitVector
refs #18

NULL -> nullptr, I'm rusty from mongo work
refs #18

fix some old apis to fit what get_key_after_bytes needs:

make keyrange_compare_s hold an FT instead of FT_HANDLE to prepare for reuse refs #18

make keyrange_compare_s hold a const dbt refs #18

fix const-correctness of fill_bfe functions refs #18

add unit test for an empty dictionary
refs #18

implement get_key_after_bytes for a single basement node, passes simple tests
refs #18

add test for multiple basement nodes
refs #18

fix usage of ft_search_t, we do need a full one after all
refs #18

check actually_skipped for correctness (even though it's an estimate)
refs #18

restructure test to be faster, and test keys before the beginning of the table
refs #18

don't try to read in the right basement node, just check the bns you have
refs #18

implement get_key_after_bytes for height > 0 trees
refs #18

return the amount skipped even if we hit the end of the table
refs #18

add inexact test for height > 0 trees, adjust constants so it's faster
refs #18

don't do I/O to bring in basement nodes just for get_key_after_bytes
refs #18

don't blindly check the first basement node, it's not guaranteed to be there anymore
refs #18

fix leak in get_key_after_bytes_unit.tdb
refs #18

add get_key_after_bytes to test_stress5
refs #18

switch to a callback API to save mallocs and memcpys
refs #18

closes #14

closes #13

only append -fimplicit-templates if -fno-implicit-templates is already there because of mysql/mariadb