Long identifier truncation improvements

[labkey-adam]

Rationale

Not heeding database-specific truncation rules for identifiers has led to several issues related to long column, table, and alias names, particularly those with non-ASCII characters. Here are two example issues that are resolved by this PR:

• Issue 52714: Problems with >63 byte at UTF-8 non-ASCII characters in field names
• Issue 52210: LKSM: Not showing fields with cyrillic characters in sample type grid

In these cases, DomainImpl.generateStorageColumnName() would call an AliasManager.decideAlias() variant that usually returned the name as the storage column name, without first making it "legal," which lead to silent truncation to 63 UTF-8 bytes on PostgreSQL. First fix attempt was to call a different method to ensure makeLegalName() was invoked when generating storage names, but this led to some undesirable behavior, such as provisioned columns named "Group" and "User" being stored as "group_" and "user_" (they're keywords on PostgreSQL), which then confused index creation (which of course should know the storage names, but they don't currently). IMO, provisioned column storage names should match the column names as much as possible (I'm less concerned about this for aliases). We're already quoting them appropriately. So the fix here is to stop using AliasManager to generate storage column names; instead, a new simple class StorageNameGenerator is now responsible for generating legal and unique storage column names. It calls the dialect-specific truncation method and uniquifies names with a suffix counter, but otherwise it leaves the characters as is. This means we'll start seeing special characters in provisioned tables' column names.

Changes

• Introduce StringUtilsLabKey.truncateStartToUtf8ByteLimit() that truncates from the right end of the string. Add tests. Simplify overly complex truncateToUtf8ByteLimit().
• Move the truncation and "make legal" methods to SqlDialect to allow these to be dialect-specific
• Implement correct truncation in PostgreSQL dialect
• Pass in a @NotNull SqlDialect to more AliasManager calls
• Introduce FallBackDialect for the unfortunate cases where a SqlDialect is not provided to AliasManager; it implements conservative truncation rules that work across all supported databases
• Eliminate useLegacyMaxLength flag
• Force callers to specify the number of characters they need reserved for suffixes, etc. Previous code made blanket assumptions that were often incorrect, unnecessary, and/or redundant.
• Switch to surrogate-pair-aware truncation so we don't end up with half characters in our names
• Widen StorageColumnName and mvIndicatorStorageColumnName to ensure even the longest generated names will fit

[labkey-jeckels]

Minor suggestions. I haven't test the code.

[labkey-adam]

Merged to #6498