Experiment: Refine ASCII column definitions

[theory]

Columns that store SHA hash hex strings or URIs can be ASCII or use "C" collation. Update each engine to use the appropriate encoding or collation for such columns.

Issues that remain:

MySQL

MySQL has trouble comparing strings of different types. On 5-5-5.6 it complains:

Illegal mix of collations (utf8mb4_unicode_ci,COERCIBLE) and (ascii_general_ci,IMPLICIT) for operation '='
DBD::MariaDB::db do failed: Illegal mix of collations (utf8mb4_unicode_ci,COERCIBLE) and (ascii_general_ci,IMPLICIT) for operation '='
Trace begun at /home/runner/work/sqitch/sqitch/lib/App/Sqitch/Role/DBIEngine.pm line 712

Cockroach also has this problem, complaining:

psql:/home/runner/work/sqitch/sqitch/lib/App/Sqitch/Engine/cockroach.sql:101: ERROR:  unsupported comparison operator: <collatedstring{en-US-u-va-posix}> = <string>
"psql" unexpectedly returned exit value 3

Oracle

Oracle requires MAX_STRING_SIZE=STANDARD, but it seems to require restarting the server, and I don't know how to do that for the Docker image.

Vertica

Has no obvious method to support column-based collation or encoding.

SQLite

SQLite supports a COLLATE keyword, but it defaults ti BINARY, which is what we'd want for SHA hash hex strings and URIs anyway. We could use NOCASE for SHA sums, but Sqitch always lowercases them anyway.

Future Work

Overall it might be better to provide a way for users to select a collation or encoding to use and then apply it.

But in the end I'm not sure it's worth the effort.