docs: correct SwinUNETR pretrained datalist count and link (#2081)

### Description
Correct the pretrained SwinUNETR tutorial's datalist count from three to
five and replace the stale `json_files` directory name with a relative
link to `datalists/`.

Validation: inspected all five JSON files at the base commit. The
training arrays contain 3, 5, 7, 12 and 24 entries, and the validation
arrays contain the same six entries. The relative link resolves locally
and `git diff --check` passes. No notebook or model was run for this
documentation-only change.

Prepared and checked with Codex assistance.

### Checks
- [x] Avoid including large-size files in the PR.
- [x] Check the changed content for sensitive information.
- [x] Use a working relative link for the repository directory.


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->

## Summary by CodeRabbit

* **Documentation**
* Updated the BTCV dataset instructions to clarify that its five JSON
files are in the `datalists` directory.

<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Signed-off-by: Lefterismpak <lefterismpak@gmail.com>
Co-authored-by: Eric Kerfoot <17726042+ericspod@users.noreply.github.com>

Author: Lefterismpak

self_supervised_pretraining/swinunetr_pretrained/README.md

@@ -32,8 +32,8 @@ Pre-training Dataset: Datasets from multiple sources has been used. Below is a l
 Fine-tuning Dataset: The dataset from Beyond the Cranial Vault Challenge
 [(BTCV)](https://www.synapse.org/#!Synapse:syn3193805/wiki/217789)
 2015 hosted at MICCAI, was used as a fully supervised fine-tuning task on the pre-trained weights. The dataset
-consists of 30 3D Volumes with annotated labels of up to 13 different organs [2]. There are 3 JSON files provided in the
-json_files directory for the dataset. They correspond to having a different number of training volumes ranging from
+consists of 30 3D Volumes with annotated labels of up to 13 different organs [2]. There are 5 JSON files provided in the
+[datalists](datalists) directory for the dataset. They correspond to having a different number of training volumes ranging from
 3, 5, 7, 12 and 24. All 5 JSON files have the same validation split.
 
 ### 2. Pretrained SwinUNETR Weights