Move AttrInfo from back-ldbm.h to attr.c

Still not usable.

MUST be defined).

one can select NDBM as NDBM is not supported.

behind -USLAPD_SCHEMA_COMPAT.

unless access requested is WRITE.  This allows you to apply
an ACL to limit search/reading of no-user-modification attributes.
Writes, of course, are always prohibited (by do_add, do_modify).

ldap_ucs2_t is a short.  The unicode routines may break if
sizeof(short) != 2...

bugs and adds temporary code to glue things until we sort
out unicode UTF-8 vs UCS-4 vs UCS-2 issues.

UTF-8 internally.

syntax.

Revert normalization to matching rule per discussions with Julio.
May need separate normalization routines for stored value and asserted
value. Currently rely on passed in syntax/mr to allow "special" behavior.
Reworked filters to pass struct berval * instead of char *. (needs work)
Validation, normalization and matching needed.

Well help when we split assertion options out from the description.

routine.  Could be combined into one routine.
Modify slapd matching rule struct to only have match function.
Modify old attribute such that 'bin' implies octetString, not 'binary'.
Add compatibility for IA5 strings.  Only directoryStrings were handled
before.
Treat attribute types without syntax as incomplete, not default.
Add OctetStringValidate (always returns okay).
Add {UTF8,IA5}StringValidate/Normalize  (IA5 based loosely on
case_exact_normalize).  Need case{Exact,Ignore}UTFMatch, using IA5 versions
for now.
Removed default of syntax/mr handlders, should just skip
registration of syntax/mr's without handlers.
Added comments to slap.h about types versus descriptions.

s/case_ignore_compare/caseIgnoreIA5Match/
s/case_exact_normalize/caseExactIA5Normalize/
s/case_exact_compare/caseExactIA5Match/

to make room for UTF-8 and T.61 varients.

Add US ASCII optimizations macros.
#ifdef out unused routines
Ready to hack getdn.c and others to support UTF-8

We may likely want to expose the routines at a later date (ldap_utf8.h?).

ldap_utf8_strcspn, ldap_utf8_strtok.