[Driver] Single-module lowering flow in driver_api.cc

[Lunderberg]

Prior to this commit, a build that used multiple targets needed to provide tvm::build with a Map<Target, IRModule> specifying which target should be used to compile each IRModule. As a result, lowering passes could not introduce new targets based on a PrimFunc's content (e.g. a with T.target() frame to delegate out to another device), nor simplify based on cross-device subroutines (e.g. simplify a host-side conditional based on the known output of a device-side internal subroutine).

This commit makes the tvm::attr::kTarget attribute ("target") be the single source of truth for where a PrimFunc will be executed. Other existing methods for specifying the target (the target parameter for tvm.build, the keys in a Map<Target,IRModule>, the parameter to the pass tir::transform::BindTarget) are still accepted as inputs, and may provide a default value for tvm::attr::kTarget if the attribute is missing, but may not overwrite the target attribute.

This is part of a series of commits to simplify the handling of multi-target builds.

[tvm-bot]

Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.

• No users to tag found in teams: driver _{See #10317 for details}

_{Generated by tvm-bot}

[Lunderberg]

This PR is a subset of the changes introduced in https://github.com/apache/tvm/pull/14862, split out for ease of testing and reviewing. It is currently a draft PR, as it depends on changes made in https://github.com/apache/tvm/pull/14944, and will be rebased onto main after https://github.com/apache/tvm/pull/14944 lands.

[Lunderberg]

Rebased onto main now that #14944 has landed, marked as ready for review.

[Lunderberg]

Converting to draft, as the fixes for the last few CI failures depend on changes in https://github.com/apache/tvm/pull/14986, and so this PR should wait until https://github.com/apache/tvm/pull/14986 has landed.

[Lunderberg]

With https://github.com/apache/tvm/pull/16184 landed, this PR should (hopefully) be able to land without requiring all low-level codegen to directly support function calls. For codegens that do not yet support function calls, device-side kernels can have subroutines inlined to avoid a recurrence of https://github.com/apache/tvm/pull/16033.

Currently, this PR is rebased on top of main, but does not use of InlinePrivateFunctions. I expect it to fail in CI for Metal codegen, and to have the same ethos-u failure modes as previously. These should be resolvable by applying InlinePrivateFunctions, conditional on the target being used.