Bump smithy to 1.49.0

[kubukoz]

[@Baccata's comment]

In addition to bumping smithy, adds an option for lenient tagged union decoding, used in the context of AWS. This fixes a few compliance tests that were previously broken (#1424)

PR Checklist (not all items are relevant to all PRs)

• Added unit-tests (for runtime code)
• Updated changelog

[kubukoz]

[error] [aws-http4s] - aws.protocoltests.ec2#QueryLists(client|Request): Ec2EmptyQueryLists 89ms
[error] [aws-http4s]    the result value: "Action=QueryLists&Version=2020-01-08&ListArg=" was not equal to the expected TestCase value "Action=QueryLists&Version=2020-01-08". (modules/tests/src/smithy4s/tests/ProtocolComplianceSuite.scala:183)
[error] [aws-http4s] 
[error] [aws-http4s]     [INFO] 23:25:07 [ProtocolComplianceSuite.scala:173] Does not serialize empty query lists.
[error] [aws-http4s] 

I think I can give this a shot, but not today.

[Baccata]

The ONLY way we can justify rolling the bump in 0.18 is by showing some confidence that it's VERY unlikely to cause runtime issues, unless people have written some very esoteric interpreters.

That's what your experiments have shown (in our DMs), with the only way you've made it blow has been with this

SomeType.hints.get(Trait).map(_.breakingChanges.map(_.map(_.severity))))

So you need to change the wording to something like:

 

In the vast majority of applications using Smithy4s, it will not cause runtime issues. However, in the unlikely event that you have custom interpreters that query the .breakingChanges field of a Trait hint, you'll have to ensure that all the libraries pulled by your application are compiled against smithy4s 0.18.19.

[kubukoz]

I'll do some more tests to increase my confidence ;)

[kubukoz]

updated the entry, does it look good?