fix: fix literal of Nullable and order of limit & offset in Clickhouse.

[Lunaticus7]

Fix two bugs of Clickhouse backend.

1. The order of limit and offset in compiled SQL is reversed 🤣.
   https://clickhouse.tech/docs/en/sql-reference/statements/select/limit/#limit-clause

2. When literal None to some Nullable column's type will raise error.

expr = table.mutate(ibis.literal(None, table['nullable_string_column'].type()).name('new_column'))
expr.compile()
# AttributeError: 'NoneType' object has no attribute 'replace'

[github-actions]

Unit Test Results

       19 files         19 suites   1h 34m 25s ⏱️
10 756 tests   8 401 ✔️   2 354 💤 1 ❌
54 461 runs  41 185 ✔️ 13 274 💤 2 ❌

For more details on these failures, see this check.

Results for commit e5c79a5.

[Lunaticus7]

CI failed, emmmm, there is an import error ?

https://github.com/ibis-project/ibis/pull/2983/checks?check_run_id=3632398102#step:11:105

[cpcloud]

@Lunaticus7 Unfortunately, we're now picking up a later version of impyla that uses a dependency that isn't declared in its setup.py. I'll patch the conda-forge package to include this dependency, hopefully I don't also need to package puresasl as well.

I've created an issue on the impyla issue tracker (cloudera/impyla#471).

[cpcloud]

I've put up a PR to add pure-sasl to the impyla conda package, which should at least allow CI to pass conda-forge/impyla-feedstock#28

[cpcloud]

Some small nits.

Other than the CI (which doesn't really matter for this PR anyway, but we should still wait for it to pass) this LGTM.

[Lunaticus7]

Closed, split this PR to #2984 #2985

cc @cpcloud