We want to pass more structures through the Bison stack instead of using Lex members.

In particular, we want to get rid of Lex->charset.

Reasons:

• The current implementation is hard to follow and change. It has caused a few bugs already. Passing data through the bison stack is much easier to follow!
• We'll have contextually typed collations soon:

  -- Use the xxx_uca1400_as_ci collation of the column implicit character set:

  CREATE TABLE t1 (a CHAR(10) COLLATE uca1400_as_ci);

  The above will resolve to:

• COLLATE utf8mb4_uca1400_as_ci (in case if the column character set is later resolved to utf8mb4)
• COLLATE utf16_uca1400_as_ci (in case if the column character set is later resolved to utf16
• and so on

So the code logic must be simlified before this change.

This change will also fix a few problems:

• MDEV-27690
• MDEV-27853
• MDEV-28067
• MDEV-28118
• MDEV-28119