WL#7737: InnoDB: Support descending indexes

This is the InnoDB part of WL#1074, adding support for descending indexes.

InnoDB will store an ASC/DESC flag in SYS_FIELDS and in dict_field_t.

Change buffering will be prohibited on indexes that contain
descending columns, or if the primary key contains descending columns.
This could be lifted with Per-index change buffer.

The InnoDB SQL parser will not make use of descending indexes.
Because of this, we will require FTS_DOC_ID_INDEX(FTS_DOC_ID)
to be an ascending index.

dict_field_t: Add the flag is_ascending.

dict_index_add_col(), dict_mem_index_add_field(): Add the flag is_ascending.

dict_load_field(): Load the ASC/DESC flag from SYS_FIELDS.POS

cmp_data_data(), cmp_dfield_dfield(): Add the flag is_asc. Those
callers that are only interested whether the result is 0 can pass any
value for is_asc.

cmp_dfield_dfield_like_prefix(): Rename to cmp_dfield_dfield_eq_prefix(),
and return a Boolean instead of a less,equal,greater status.

dtuple_coll_cmp(): Rename to dtuple_coll_eq(), and return a Boolean
instead of a less,equal,greater status.

cmp_dtuple_rec(), cmp_dtuple_rec_with_match_low(),
cmp_dtuple_rec_with_match(), cmp_dtuple_is_prefix_of_rec(): Add the
parameter dict_index_t, so that we will know which index fields are
ascending and which ones are descending.

cmp_decimal(), cmp_whole_field(), cmp_data(), cmp_data_data(): Add the
parameter is_asc.

row_ins_duplicate_online(): Add the parameter dict_index_t.

row_merge_tuple_cmp(), row_merge_tuple_sort(): Replace the parameters
n_uniq, n_field with index.

page_find_rec_max_not_deleted(): Rename to
page_find_rec_last_not_deleted(). The last record in a descending
index would be the minimum value, not the maximum.

row_search_get_max_rec(): Make the function public. Support descending
indexes as well.

dict_index_is_asc(): Check if an index consists entirely of ascending
fields.

opt_calc_index_goodness(): Ignore indexes that are not entirely ascending,
and assert that descending indexes are not being used.
This is only used in the InnoDB internal SQL parser.

ibuf_should_try(): Deny change buffering unless dict_index_is_asc().
This can be re-allowed when implementing per-index change buffer.

fts_get_max_doc_id(): Use row_search_get_max_rec().

innobase_table_flags(): Check that there is only one column in
FTS_DOC_ID_INDEX. We already had a similar check in ADD FULLTEXT INDEX,
but the check in CREATE TABLE was more sloppy.

innobase_table_flags(): Check that DESC is not specified on FTS_DOC_ID
during CREATE TABLE.
  
innobase_fts_check_doc_id_index(): Check that the FTS_DOC_ID_INDEX on
FTS_DOC_ID is ascending, not descending.