Remove automatic padding for fixed types as it should be done by custom serialization strategy

[Chuckame]

Is your feature request related to a problem? Please describe.
Currently, when a byte-array or string is encoded as fixed, it tries to match the fixed size by padding with 0s at the beginning (for ByteArray and ByteBuffer) or at the end (for String), which ends up to a modified content.

This should be done on the user side, depending on its needs.

Describe the solution you'd like
Totally remove the automatic-padding feature, and fail when the given string/ByteArray/ByteBuffer is not of the same size as the expected fixed schema.

To still be able to pad with 0s, the user has to provide its own AvroSerializer, providing explicitly the fixed schema and doing its padding logic inside serializeAvro and deserializeAvro. Here is an example:

public object ZeroPaddedByteArraySerializer: AvroSerializer<ByteArray>("zero-padded-byte-array") {
    override fun getSchema(context: SchemaSupplierContext): Schema {
        return context.fixed?.createSchema() ?: throw SerializationException("ZeroPaddedByteArraySerializer requires @AvroFixed annotation")
    }

    override fun serializeAvro(encoder: AvroEncoder, value: ByteArray) {
        // pad with 0s to match 15 bytes
        encoder.encodeFixed(value.zeroPadded(15, endPadded = true))
    }

    override fun deserializeAvro(decoder: AvroDecoder): ByteArray {
        TODO("Not yet implemented")
    }
}

[Chuckame]

Released in v2.0.0