Reland "HttpStreamPool: Group TCP attempts in slots"

This is a reland of commit b6b720b7be6b6951c37bb2cc38c231dc9b35539f.
The original CL was reverted because it caused CHECK() failures.

Diff from the original CL are:
* Don't use `NonSlowTcpBasedAttemptCount()` to calculate active stream
  counts. It didn't make sense.
* Add `CanStartFallbackTcpBasedAttempt()`, which returns true if there
  is a slow attempt for one address family (e.g. IPv4) and no
  corresponding attempt for the other address family (e.g. IPv6) has
  been started yet.
* Change a CHECK() with DumpWithoutCrashing() in
  `CreateTextBasedStreamAndNotify()` to investigate potential bugs.
* Add a test to ensure the previous crash doesn't happen.

Original change's description:
> Revert "HttpStreamPool: Group TCP attempts in slots"
>
> This reverts commit ef6ba930ba83b74518f3ed7d89247c4c82e84764.
>
> Reason for revert:
> The Suspect CL https://chromium-review.googlesource.com/q/Iba4feb76ecbb3779cbfc80ffd0d3d311b24328ad is highly likely the cause of the crash. Please see the crash report http://crash/58cb8b91483394cf.
> Rebase the Revert CL on top of main branch with conflicts allowed option to prevent false positive merge conflicts.
>
> If you agree with the revert CL, please approve it and mark as Commit-Queue +2 to submit it.
>
> More details can be found in the go/crash-revert-cls-creation.
>
> Original change's description:
> > HttpStreamPool: Group TCP attempts in slots
> >
> > This CL introduces the concept of a TcpBasedAttemptSlot in the
> > HttpStreamPool. The purpose of this slot is to group a single IPv4 and a
> > single IPv6 connection attempt together. This allows the pool to treat
> > them as a single unit for the purpose of connection limits.
> >
> > This change is motivated by performance issues on dual-stack networks
> > where one address family (e.g., IPv6) might be slow or unresponsive. In
> > such scenarios, the slow attempt could occupy a connection slot for an
> > extended period, preventing other, potentially faster, attempts from
> > being made.
> >
> > By grouping the attempts, the HttpStreamPool can more intelligently
> > manage its resources and avoid getting stalled by a single slow address
> > family.
> >
> > This behavior is similar to the non-HEv3 code path, where
> > TransportConnectJob could have both IPv4 and IPv6 SubJobs at the same
> > time.
> >
> > Another approach considered is to count attempts per address family. WIP
> > CL is [1]. Downsides of the approach are:
> > * AttemptManager::CanAttemptConnection() can be called without addresses
> >   so we can't count per address family.
> > * There is no single concept of the group being "full". The CL uses
> >   NonSlowTcpBasedAttemptCount() to proxy it, but it breaks some
> >   assumptions the rest of the code has. For example, The CL is failing
> >   on a test because AttemptManager creates a HttpBasicStream exceeding
> >   the group's limit. To avoid this, we need to cancel attempts in the
> >   other address family when an attempt succeeds.
> > * The WIP CL always increment the pool's total connecting count when
> >   creating an attempt. This means that a group could have up to 12
> >   connecting attempts. This could make Chrome to reach the pool's limit
> >   (256) more often than the non-HEv3 code path.
> >
> > [1] https://crrev.com/c/6852978
> >
> > Change-Id: Iba4feb76ecbb3779cbfc80ffd0d3d311b24328ad
> > Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/6968320
> > Reviewed-by: Adam Rice <ricea@chromium.org>
> > Commit-Queue: Kenichi Ishibashi <bashi@chromium.org>
> > Cr-Commit-Position: refs/heads/main@{#1521020}
> >
>
> Change-Id: I3d526deb06da385185c839896d127aee91d4d1b4
> Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/6991510
> Reviewed-by: Kenichi Ishibashi <bashi@chromium.org>
> Reviewed-by: mmenke <mmenke@chromium.org>
> Commit-Queue: Kenichi Ishibashi <bashi@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#1521716}

Bug: 383606724
Change-Id: Iaeff5aeada5f7ae0ed5b980335903de01eb693f1
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/7006044
Reviewed-by: Adam Rice <ricea@chromium.org>
Commit-Queue: Kenichi Ishibashi <bashi@chromium.org>
Cr-Commit-Position: refs/heads/main@{#1524106}