Performance tuning

[grigi]

tortoise.models.__init__ restructure give 25-30% speedup in select benchmark.
Usage of pypika>=0.15.6 gives 53% speedup in insert benchmark.

I also cleared up a few lint warnings.
As per findings in https://github.com/tortoise/orm-benchmarks
(The benchmark is currently only using SQLite, but both these changes should be generic)

[coveralls]

Pull Request Test Coverage Report for Build 201

• 31 of 40 (77.5%) changed or added relevant lines in 9 files are covered.
• 310 unchanged lines in 11 files lost coverage.
• Overall coverage decreased (-0.4%) to 90.313%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
tortoise/backends/base/client.py	1	2	50.0%
tortoise/models.py	14	16	87.5%
tortoise/backends/mysql/client.py	0	2	0.0%
tortoise/backends/asyncpg/client.py	0	4	0.0%

Files with Coverage Reduction	New Missed Lines	%
tortoise/backends/base/executor.py	1	95.56%
tortoise/backends/mysql/init.py	2	100.0%
tortoise/models.py	2	94.06%
tortoise/backends/asyncpg/init.py	2	100.0%
tortoise/backends/base/client.py	7	91.07%
tortoise/backends/asyncpg/schema_generator.py	8	100.0%
tortoise/backends/mysql/schema_generator.py	13	100.0%
tortoise/backends/asyncpg/executor.py	13	100.0%
tortoise/backends/mysql/executor.py	29	100.0%
tortoise/backends/asyncpg/client.py	113	80.37%

Totals
Change from base Build 191:	-0.4%
Covered Lines:	1859
Relevant Lines:	2001

---

💛 - Coveralls

[abondar]

What is the case here? You managed to catch it without transaction?

[grigi]

That was because the linter (correctly) complained that it was possible for self.transaction to not exist. Which is now forcefully set to None instead. AttributeError is if it is a None value.

I should probably make it clearer? Or is it perferred for it to raise a NPE?

Now that I think of it, a NPE would point out that something happened that should never be able to happen...