Diagnosing low performance

**What steps did you take:**

I have a fairly small ytt configuration, with 36 files and 1600 lines. Among these, 20 files are values (totaling 1100 lines).
```
-------------------------------------------------------------------------------
Language files blank comment code
-------------------------------------------------------------------------------
YAML 36 48 289 1618
```

The templates (which I unfortunately cannot share) do not do anything I would consider heavy. They just generate manifests for regular Kubernetes resources.

**What happened:**
It takes 3.9 seconds to generate the 5700 lines of output on a fairly beefy Lenovo P1 machine.

I tried to remove all values files and add them back one by one, and one can see the generation time dramatically increasing:
```
381ms 748µs 771ns
421ms 23µs 781ns
556ms 209µs 636ns
625ms 745µs 631ns
804ms 283µs 465ns
811ms 94µs 171ns
917ms 907µs 756ns
1sec 33ms 792µs 286ns
1sec 142ms 491µs 30ns
1sec 768ms 729µs 36ns
2sec 135ms 703µs 219ns
3sec 869ms 202µs 205ns
```

**What did you expect:**
With a fairly small configuration, I would expect the generation to be almost instantaneous.

**Anything else you would like to add:**
Is there a way to diagnose where ytt spends the generation time?
The `--debug` flag is generating a lot of lines that are difficult to inspect.

**Environment:**

- ytt version (use `ytt --version`): 0.49.0
- OS (e.g. from `/etc/os-release`): Manjaro Linux

---
Vote on this request

This is an invitation to the community to vote on issues, to help us prioritize our backlog. Use the "smiley face" up to the right of this comment to vote.

👍 "I would like to see this addressed as soon as possible"
👎 "There are other more important things to focus on right now"

We are also happy to receive and review Pull Requests if you want to help working on this issue.

Start by reproducing the reported 3.9-second generation time with a comparable ytt configuration and inspect the output from the `--debug` flag. The issue names no source files or tests, so identifying the relevant execution path will require reading the ytt CLI and profiling or benchmarking its generation process. Done means locating the dominant cost and documenting a practical way to diagnose it.

Written by the indexing model from the issue text.