[#588] First step in support for PGAdmin client; [#522] First step in array support

This commit takes the first of many steps towards supporting the full functionality required by #588 and #522.

Previously, Octo did not support ARRAY() syntax, leading to breakage during the PGAdmin connection process (#588). This is now fixed so that this process can continue. Note that this implementation of arrays does not satisfy the requirements of #522, as only one form of array syntax is supported with these changes.

This was accomplished by creating new SqlArray statement type and LP_ARRAY logical plan type and adding logic for handling these new types to the various modules accordingly that recurse through the parse tree and logical plans. Similarly a new ARRAY token was added to the parser and lexer, and a new array_constructor rule was added to the parser.

To make the new syntax function, a new GetScalarOrArray M helper routine was added to replace the previous GetScalar helper function. The reason for this is that ARRAY() is only used for handling subqueries, which is the same scenario in which GetScalar is used. This required adding a new to_array member to the LpExtraDetail struct for use in tmpl_print_expression, the caller of GetScalar and now of GetScalarOrArray.

Related to this, generate_physical_plan was updated to conditionally omit setting stash_columns_in_keys to TRUE when processing an ARRAY() call.

Additionally, the new ARRAY() constructor syntax was added to the documentation. Similarly, a new tests/test_array_syntax.bats.in test file was added to validate the behavior of the new syntax.

Note that a NullDereference clang warning from the new array_STATEMENT case in lp_generate_where.c was added to tools/ci/clang_tidy_warnings-release.ref as this case is guarded by asserts, but clang does not see this in release builds.

Finally, various outrefs were updated to account for the new GetScalarOrArray helper function in plan names, as well the consolidation of various pg_* functions and revision of the default PostgreSQL encoding mentioned below.

Various miscellaneous fixes and enhancements were also made:

• Added loop to handle_query.c to allow for handling of Query messages containing multiple queries (previously only the first query would be processed)
• Added current_database, pg_encoding_to_char, pg_is_in_recovery, pg_is_xlog_replay_paused, and has_database_privilege SQL functions to seed data (needed for PGAdmin startup process)
• Changed default PostgreSQL encoding to ASCII for compatibility with PGAdmin defaults (Note: the is the default encoding in PostgreSQL. This change caused no breakage, so it's not clear why UTF-8 was chosen previously).
• Stubbed out pg_settings, pg_database, pg_roles, and pg_user tables in octo-seed.sql (needed for PGAdmin startup process)
• Fixed formatting/style of comparison in readline_get_more.c
• Consolidated pg_catalog functions into one new M file, _ydboctopgfunctions.m, and removed previously scattered routines
• Added new OCTOLIT_ONEROWTABLE to replace hardcoded literal
• Revised lexer/parser to accept special functions without parenthesis in function calls and added these functions to the seed data and _ydboctopgfunctions
• Updates to handle_password_message and related tests to issue valid (though unused by Octo) information to PGAdmin:
  • Store username on %ydboctosession for reference by various SQL user information functions
  • Omit session_id when storing username (inline comments explain this)
  • Revised callers of handle_password_message to omit session parameter
• Updated pg_User M routine to lookup username and added comment, so that this function actually returns information to PGAdmin (or other clients)
• Added new test_pgadmin.bats.in test file to validate various changes related to PGAdmin support
• [#435 (closed)] Removed useless parser rules for SET statements
• Revised case of return value of current_schema function for consistency with uppercase identifiers
• Removed now unneeded __wrap_ydb_set_s call in `test_handle_password_message
• Added column name case conversion to Rocto row descriptions for compliance with PostgreSQL clients
  • Updated various tests (test_make_row_description and test_select_columns_psql.bats.in) and outrefs accordingly
• Moved _ydboctopgTableIsVisible.m and _ydboctopgGetUserbyid.m routines into _ydboctopgfunctions.m and updated octo-seed.sql accordingly
• Updated test_cancel_request.bats to increase -global_buffers to prevent TRANS2BIG failure in TCR03
• Updated test_cancel_request.bats to increase -global_buffers to prevent TRANS2BIG failure in TCR03
• Added CONCAT function to docs
• Added TOLOWER, TOUPPER, TOUPPER_STR, and TOLOWER_STR macros to reduce duplication when converting between string cases
• Revised find_column_alias_name to retrieve table_alias_STATEMENTs in case of array_STATEMENTs; updated outref
• Added sed conversion for session user in TMU02
• Added more user friendly comment in case of build.sh failure outside of BATS tests
• Added omitted initialization of use_valgrind in build.sh, which was previously uninitialized