fix: Don't depend on Serde to provide fields in the right order

[Kriskras99]

This is a (failing) test for when the field order differs between the struct (Serde) and the schema:

The application panicked (crashed).
  assertion `left == right` failed
    left: Foo { a: "World", b: "Hello" }
   right: Foo { a: "Hello", b: "World" }
in avro/src/ser_schema.rs, line 2940
thread: ser_schema::tests::different_field_order_serde_vs_schema

There are two options:

• Require that the field order is the same
• Modify ser_schema.rs (and potentially de.rs) to care about the field order

[martin-g]

Please fix the formatting issue.

[Kriskras99]

Fixed, CI will still fail as the test does fail and we need to decide what the right course of action is.

[martin-g]

we need to decide what the right course of action is

I won't have time to debug it in the next days but I was going to bet that we probably don't use the lookup map but we do -

avro-rs/avro/src/ser_schema.rs

Lines 298 to 299 in 335e74e

	.lookup
	.get(key)

So, it must be something else.

[Kriskras99]

The lookup is only used to find the field information from the name.
After that, the field is just written as there is no buffering for unordered fields.
I can write a fix for that, or we can decide that the Serde field order must match the Schema field order (and add that to the docs).

[Kriskras99]

I've added support for out-of-order fields in the serializer. In the deserializer no changed are necessary.
This changes test introduced in #227 as an #[avro(default = ...)] is now required when using any of the Serde skip attributes. I'm also working on a PR to check the Serde attributes in avro-derive (and add documentation for the attributes avro-derive supports as they don't seem to be documented anywhere).

[Kriskras99]

I've downgraded it to a debug_assert and added an extra check in serialize_next_field so that it should never trigger it all.
I do think we should keep the debug_assert, because it has no performance impact but if the situation does happen it prevents data corruption.

[Kriskras99]

The match self.field_position.cmp(&field.position) was suggested by clippy, please let me know if you don't like it. Then I'll revert to if/else if/else and add an #[allow].