Symbolic names: Generate dependsOn for "existing" resources

[anthony-c-martin]

Closes #10343

[github-actions]

Test Results (linux-x64)

       33 files  ±0         33 suites  ±0   35m 12s ⏱️ - 4m 24s
10 398 tests +1  10 362 ✔️ +1  36 💤 ±0  0 ❌ ±0 
12 615 runs  +1  12 579 ✔️ +1  36 💤 ±0  0 ❌ ±0 

Results for commit f3c13f5. ± Comparison against base commit 6c3a71d.

[github-actions]

Test Results (linux-musl-x64)

       33 files  ±0         33 suites  ±0   37m 6s ⏱️ + 1m 24s
10 398 tests +1  10 362 ✔️ +1  36 💤 ±0  0 ❌ ±0 
12 615 runs  +1  12 579 ✔️ +1  36 💤 ±0  0 ❌ ±0 

Results for commit f3c13f5. ± Comparison against base commit 6c3a71d.

[github-actions]

Test Results (win-x64)

       33 files  ±0         33 suites  ±0   32m 52s ⏱️ - 3m 48s
10 410 tests +1  10 374 ✔️ +1  36 💤 ±0  0 ❌ ±0 
12 626 runs  +1  12 590 ✔️ +1  36 💤 ±0  0 ❌ ±0 

Results for commit f3c13f5. ± Comparison against base commit 6c3a71d.

[github-actions]

Test Results (osx-x64)

       33 files  ±0         33 suites  ±0   1h 39m 3s ⏱️ + 11m 9s
10 402 tests +1  10 366 ✔️ +1  36 💤 ±0  0 ❌ ±0 
12 619 runs  +1  12 583 ✔️ +1  36 💤 ±0  0 ❌ ±0 

Results for commit f3c13f5. ± Comparison against base commit 6c3a71d.

[jeskew]

I believe this isn't strictly necessary since a reference function that refers to a resource by symbolic name and does not specify an API version creates an implicit dependency in ARM, but this does give us the flexibility to start using API versions in symbolic name reference functions if so desired.

[anthony-c-martin]

I believe this isn't strictly necessary since a reference function that refers to a resource by symbolic name and does not specify an API version creates an implicit dependency in ARM, but this does give us the flexibility to start using API versions in symbolic name reference functions if so desired.

That's fair. In this case it wasn't a conscious change - I had intended to do this, but the code I'd written didn't behave in the way I intended it to (even if it happened to work!). So here I just wanted to bring the actual behavior in-line with my own mental model.

We could definitely revise the approach in the future if we feel like dependsOn isn't necessary at all - I'd just like that to be a conscious decision.